Method and system for managing a group of computers6212581Abstract The system and method of this invention automatically manages a group of computers by automatically gathering data, storing the data, analyzing the stored data to identify specified conditions, and initiating automated actions to respond to the detected conditions. The invention, hereafter "SYSTEMWatch AI-L", comprises a SYSTEMWatch AI-L client which turns a computer into a managed computer, a SYSTEMWatch AI-L console, which turns a computer into a monitoring computer, a SYSTEMWatch AI-L send facility, which allows a system administrator to send commands to various SYSTEMWatch AI-L clients through the SYSTEMWatch AI-L console, and a SYSTEMWatch AI-L report facility which allows a system administrator to query information collected and processed by the SYSTEMWatch AI-L clients and SYSTEMWatch AI-L consoles. Claims I claim: Description FIELD OF THE INVENTION
TABLE 1
FEATURE DESCRITPION
NAME A property must have a name.
TYPE A property must have a type, which corresponds to the type
of the data to
be stored in the field.
FORMAT A property may optionally have a string which describes how
the data in
the field should be formatted. The format string is similar
to the C
language's printf()'s formatting control.
HEADER A property may optionally contain a string which will be
displayed as the
column header when a report featuring records containing the
property is
displayed.
DISPLAYUNIT A string used by the reporting facility which is appended to
the data in the
field during a report. Thus, if the PROPERTY is a
description of memory
utilization in kilobytes, an appropriate DISPLAYUNIT might
be "kb"
DISPLAYTYPE Some display formats are commonly used through SYSTEMWatch
AI-L.
DISPLAYTYPES are keywords which corresponds to a particular
FORMAT. Examples of DISPLAYTYPEs include STRING20, for a
string
limited to 20 characters in width, DATESMALL, for displaying
date in
mm/dd format, PERCENT, for automatically display numbers
between
0.0 and 1.0 as percentages (e.g.: 0.52 is displayed as 52%)
SHORTDESC A PROPERTY may optionally contain an abbreviated description
of the
PROPERTY.
LONGDESC A PROPERTY may optionally contain a long description of the
PROPERTY.
2. ENTITY Conceptually, ENTITYs are similar to database tables. In SYSTEMWatch AI-L, ENTITYs are used to group related PROPERTYs. FIG. 7 illustrates the concept that each piece of data in the database is associated with a given PROPERTY and a given ENTITY. In this document, it will be necessary to refer to certain combinations of ENTITYs and PROPERTYs. The construction <entity name>_<property name> (e.g.: IGNORE_IGNORETIME) will refer to a database entry with an entity equal to <entity name> and a property equal to <property name>. In addition to ENTITYs and PROPERTYs, the database, 41, in SYSTEMWatch AI-L also has these additional features: 1. Host Information Each piece of data in database, 41, automatically has host information associated with it. Thus, as data is stored in the database, the database automatically associates the host from which the data originated from. This is because in SYSTEMWatch AI-L, data is "owned"by the host from where the data originated. Other hosts may request a copy of the data since SYSTEMWatch AI-L has communications capabilities. Some data may be stored in a central location (e.g.: a SYSTEMWatch AI-L console) if it is relevant to multiple computers. Because each piece of data has host information associated with it, a SYSTEMWatch AI-L console can conslidate data from multiple hosts. 2. Time Information Each piece of data in database, 41, has a time field associated with it. The time field by default has the last time the data was updated, but SYSTEMWatch AI-L provides a mechanism of changing the time field so its possible to store some other time in the field. 3. Name Each piece of data in database, 41, has a key field which is called the name field. A name field must be unique for a given ENTITY, PROPERTY, and host (the name of a computer). Thus, within an ENTITY and PROPERTY used for tracking computer processes, the name field might be the process id since process ids are unique on each computer, so by specifying the ENTITY name, PROPERTY name, and host name, the name field forms a unique key to locate the data. 4. Value Of course, a database stores data. In SYSTEMWatch AI-L, the term value refers to the data stored in the database. In one example, database, 41, is currently implemented as a relational database: One table is used for describing ENTITYs. This table is used to associate ENTITYs with PROPERTYs. Another table is used for describing PROPERTYs. Finally, another table holds the information, which can be located by providing an ENTITY name, PROPERTY name, and the name field of the data. This table also contains the associated host and time information. In another embodiment, database, 41, can also be implemented with a database which is object oriented, i.e, a database which supports the ability to inherit data and methods from super and sub classes. Additional requirements of database, 41, used in the core is that the database must support certain query operations and certain set operations. Specifically, the query operations supported by the database include: 1. regular expression matching in queries. 2. creation time or update time query, i.e., searching for a data item based upon the time the data was stored in the database or based on the time the data was last updated in the database. 3. host of origin in queries, i.e., searching for a data item based on the host which created the data. 4. time comparison query, i.e., searching for data based upon a time comparison. Note: SYSTEMWatch AI-L stores its time in a manner similar to the UNIX operating system. That is to say, all time is converted to seconds elapsed since the beginning of UNDO time. The advantages of using this method is that time comparisons are easily made, and a time plus an interval can be added to obtain a future time. The set operations which database, 41, supports include: 1. set intersections (ANDs)--given 2 or more sets of data, return the elements present in all sets. 2. set union (ORs)--given 2 or more sets of data, return the elements in all sets. 3. set exclusion (NOTs)--given a first set and a second set, return elements in the first set which are not elements of the second set. Core Layer Description--The Expert System The second element of the core layer is an expert system, 40, which is used for problem detection and action initiation. The expert system, 40, is a forward chaining rule based expert system using a rule specificity algorithm. When SYSTEMWatch AI-L client, 13, is started, the expert system contains no rules. Rules are declared and incorporated into the core layer. Rules support both the IF-THEN rules as well as IF-THEN-ELSE rules. The rules used in SYSTEMWatch AI-L permit assignments and function calls within the condition of the rule. Additionally, SYSTEMWatch AI-L expert system, 40, also has the following features: a. Rules can declare variables. All variables declared within a rule are static variables. b. Rules can have an initialization section. The initialization section contains actions which must be performed only once, and before the rule is ever tested. It can, for example, contain a state declaration and an interval declaration (states and intervals are described below). It may contain variable declarations for variables used by the rules, and it may contain code to do a variety of actions. c. Rules can have, for instance, an INTERVAL and a LASTCHECK time. In accordance with the principles of the present invention, in order for a rule to be eligible for testing by the expert system, at the time of testing the clock time must be equal to or greater than the LASTCHECK time plus the INTERVAL time. The LASTCHECK time for each rule is set to the clock time whenever a rule is actually tested. This way, the INTERVAL specifies the minimum amount of time which must elapse since the last time a rule was checked before the rule becomes eligible for testing again. d. The expert system and its rules have a state property. One example of the possible states is described below. Under expert system, 40, in order for a rule to be eligible for testing, the rule's state must equal the expert system's state. All rules except one must declare a state for the rule in its initialization section. The one rule without such a declaration is a rule used by expert system, 40, to switch it into the DATA state. Other rules are responsible for managing the transition from DATA to DATA2, and from DATA2 to EXCEPT. These states are described below: DATA: The data state is assigned to rules which gather raw data from the computer system. Examples of such rules would be rules which gather the amount of free space remaining on a file system, or the amount of CPU time consumed by a process. SYSTEMWatch AI-L contains a series of rules responsible for switching states, and those rules ensures that rules with the DATA state are eligible to be tested before rules with a DATA2 or EXCEPT state. DATA2: Sometimes, a rule which performs problem detection or a rule which initiates an action requires data which can only be computed after certain raw data is gathered in the DATA state. Although the rule can compute that information directly, if that computation is necessary for a variety of rules, it is more efficient to ensure that the computation is performed only once. The DATA2 state is assigned to rules which perform this intermediate level calculation. The rules responsible for switching states ensure that DATA2 state rules are eligible for testing after DATA state rules, but before EXCEPT state rules. EXCEPT: The EXCEPT state is assigned to the remaining rules, which are used to perform problem detection and action initiation. The rule responsible for switching states ensures that EXCEPT state rules are eligible for testing only after both DATA and DATA2 state rules are tested. However, after the EXCEPT state rules are tested, the state is reset to the DATA state, and the cycle resumes. e. Each rule in the expert system also has a ONCE property. ONCE defaults to true, but can be set to false on a per rule basis by making the appropriate declarations in the initialization section of the rule, or the THEN clause, or the ELSE clause of the rule. In SYSTEMWatch AI-L, a rule is not eligible for testing by the expert system if the ONCE property for the rule is true, and if during this pass through the expert system, the rule has previously been tested true. After all the rules are declared, the expert system is in a state where it is ready to test rules. SYSTEMWatch AI-L forces the expert system component of the core layer to run through its rules whenever the execRules function is called. As described later, the SYSTEMWatch AI-L client, 13, and SYSTEMWatch AI-L console, 21, each call a execRules function in their main loop. As shown in FIGS. 8a-8b in one embodiment, the expert system functions as follows: First, if the rules have not been sorted, INQURY 59, "Have the rules been sorted?", the expert system reorders the rules by sorting them in specificity order, STEP 60. Rules are ranked in their order of specificity, with the most specific rules ordered before the least specific rules. Specificity is the total number of comparison operators (less than, less than or equal to, equal to, greater than, greater than or equal to, not equal to) and logical operators (AND, OR, NOT) contained within the boolean expression used as the test in the rules. For example, consider these boolean expressions:
TABLE 2
Boolean Expression Specificity
A AND NOT B OR (C == D) 4
(A == B) && NOT C 3
(A == B) && C 2
A == B 1
TRUE 0
If during the sorting, a group of rules has the same specificity, that group is sorted in declaration order, with the earlier declared rule ordered before a later declared rule. The reordering of the rules is only done once, during the first time the execRules function is called. If this is the first time the execRules function is called, INQUIRY 61, "Have the rules been initialized?", the expert system also performs rule initialization by running through each rule in order, and each rule's LASTCHECK time is set to zero, ONCE state is set to TRUE, and any statements contained in the rule's initialization section are executed, STEP 62. Subsequent to initialization or if initialization was previously performed, expert system, 40, begins testing rules in sequence. First, the expert system sets its rule state to a empty string, STEP 63. Then the expert system sets its current rule pointer to the first rule, STEP 64. It makes the current rule be the rule pointed to by the current rule pointer, STEP 65. Then, before testing the rule, the expert system checks to see if the current time is greater than the rule's LASTCHECK time plus the rule's INTERVAL time, INQUIRY 66. If so, the required interval has elapsed, and the rule is not disqualified from testing. Otherwise the rule is disqualified from testing during this pass through the rules. If the above inquiry is affirmative, INQUIRY 67, the expert system checks to see if the expert system rule state is equal to the current rule's state. If they are equal, the rule is not disqualified from testing during this pass through the rule. Otherwise the rule is disqualified. Should the expert system rule state equal the current rule's rule state, the expert system checks to see if the rule's ONCE variable is set to TRUE, INQUIRY 68. If it is, and if this rule has ever tested TRUE during the current call to the execRules function. If so, the rule is disqualified from testing during this pass through the expert system. If not, the rule is eligible for testing. If a rule is eligible for testing, the expert system tests its condition and sets the rule's LASTCHECK time to be equal to the current time, STEP 69(The rule's LASTCHECK time is updated when the condition is tested.). If the condition is true, the expert system then executes the THEN clause of the rule, STEP 70. If the condition is false, the expert system executes the ELSE clause of the rule, STEP 71, if it exists. What happens next depends upon what happened during the rule qualification and rule testing state. If the rule was disqualified from testing, or if the rule was tested and the condition was false, the expert system checks to see if the current rule is the last rule in the expert system, INQUIRY 72. If so, the expert system pass is completed for the time being, and the execRules function returns, STEP 74. If not, the expert system sets the current rule pointer to the next rule, STEP 73, and begins the process of checking rule testing eligibility and rule checking again, STEP 75. On the other hand, if the rule was tested, and the condition was true, then the expert system sets the current rule pointer to the first rule in the expert system, STEP 64, and the expert system begins the process of checking rule testing eligibility and rule checking again, STEP 65. Core Layer Description--Language Interperter Returning to FIG. 6, the third element of the core layer is a mechanism for configuring and controlling the database and the expert system. One preferred embodiment of this layer is an interpreter, 39, for a high level language, said language containing a mechanism of expressing database operations, database data definitions, and expert system rules. Core Layer Description--Communications Mechanism Finally, the fourth element of the core layer is communications mechanism, 42. The communication mechanism, 42, used by SYSTEMWatch AI-L is based on mailboxes. Each module has its mailboxes which is used to receive incoming data and commands. In one example, SYSTEMWatch AI-L contains two modules, SYSTEMWatch AI-L client, 13, and SYSTEMWatch AI-L console, 21. Messages are sent by deliverying files to desired module's mailbox. If the desired module is on a different computer, the delivery mechanism must be able to transport a message from one computer to another. In one example of a preferred embodiment, the communication mechanism, 42, operates by running a communications daemon on each machine which has either SYSTEMWatch AI-L client, 13, or SYSTEMWatch AI-L console, 21. A sending module delivers its message to a receiving module by passing the message to the communications daemon located on the machine where the sending module is located. Message passing is accomplished by sending messages on a TCP/IP based network using network sockets. That communications daemon then transmits the message over a computer network to the communications daemon where the receiving module is located. The communications daemon on the machine where the receiving module is located then places the message in a file in the mailbox of the receiving module. In another example of a preferred embodiment, the communications mechanism, 42, operates by placing all mailboxes of all modules in a central location, say a certain directory on a file server. On each machine which contains either SYSTEMWatch AI-L client, 13, or SYSTEMWatch AI-L console, 21, the file server directory where the mailboxes are located is made accessible. Thus, a sending module delivers its message to a receiving module simply by writing a file into the appropriate mailbox. Now that the client program organization has been explained, its possible to understand how the SYSTEMWatch AI-L client operates within the context of its bifurcated layers, i.e., the core and application layers. When the SYSTEMWatch AI-L client first begins, it consists of the core layer program reading a file containing a program written in the high level language which can be interpreted by the core. That program, including the 14 programs which that program will read, comprises the application layer for SYSTEMWatch AI-L client, 13. At this point, the database in the core layer has no data record definitions, and no data records. Similarly, the expert system within the core layer has no rules, variables, or routines. As the language interpreter, 39, portion of the core begins to interpret and execute the program, the first thing the program causes the core to do is to perform some housekeeping work. This work consists of ensuring that the communications mailboxes used by the SYSTEMWatch AI-L client are set up. After the housekeeping is done, SYSTEMWatch AI-L client, 13, causes the core to read in a series of files. These files are also files with programs written in the high level language. As each file is read, the routines, data record definitions, and rules expressed in each file are incorporated into the database, expert system, and language interpreter, 39, of the core. One preferred embodiment is to split these programs into 14 parts, consisting of the following files: 1. worksets 2. configs 3. events 4. requests 5. coms 6. lib 7. alerts 8. filesys 9. files 10. swap 11. process 12. daemon 13. actions 14. ruleinit Note that if the system administrator wanted to add additional modules to detect, analyze, and respond to additional problems, he need only write a program in the high level language conforming to convention used in the other files in SYSTEMWatch AI-L and modify the application layer to read in his program(s) before the SYSTEMWatch AI-L client reads the ruleinit program. Each of the 14 files read by the SYSTEMWatch AI-L client will now be described in detail: 1. worksets: A program which contains database declarations and routines relating to worksets. The worksets program does not declare any rules. A workset is a SYSTEMWatch AI-L ENTITY which is used to track groups of items for inclusion and exclusion, typically for including/excluding certain objects from being tested by the rules.
TABLE 3
ENTITY PROPERTY TYPE DESCRIPTION
WORKSET ITEMLIST string Actual list of colon separ-
ated items for maintaining
working sets
WORKSET ADDEL string Contains the string ADD in
case of a temporary addi-
tion record, and the string
DEL in the case of a
temporary deletion record.
An empty string means this
record is a permanent work
set record. Other values
are illegal.
WORKSET WORKSETNAME string The name of the workset
that a temporary add/delete
transaction references
WORKSET TIMEOUT integer Specifies the time at which
a temporary ADD action
will delete an item, or at
which a temporary DEL
action will ADD an item
back to the database.
0 identifies a permanent
working set record.
The routines declared in the workset program are the following:
TABLE 4
NAME FUNCTION
addItem Takes a string and adds it to a workset if the string
is not already a member of workset. Accepts the
string and a workset name.
addWorkSet Adds a string of colon delimited items to a workset.
If the workset does not exist, it is created. Can
optionally accept a time out value, which if present
means the addition is temporary, and will be deleted
from the specified workset after the timeout period
has expired. Accepts a hostname, workset name, a
string, and optionally, a time out period.
checkInclExcl Determines whether an item is on the include or
exclude list of a particular workset checkInclExcl
first checks the workset for an include list. If an
include list exists, and if the item is on the include
list, then checkInclExcl returns the string
"INCLUDE." If an include list exists and the item is
not on the include list, checkInclExcl returns an
empty string.
If an include list does not exist, checkInclExcl then
checks to see if an exclude list exists. If an exclude
list exist and the item is not on the exclude list,
checkInclExcl returns "INCLUDE". If an exclude
list exists and the item is on the exclude list,
checkInclExcl returns an empty string.
checkInclExcl accepts a workset name and an
itemstring.
checkWorkSet Reviews workset list of items waiting to be added or
deleted and executes the adds and deletes if the
appropriate time has arrived. checkWorkSet does
not accept any parameters.
delItem Accepts an itemstring and a workset, goes through
the workset and deletes every item in the itemstring
from the workset, and then returns the (modified)
workset.
delWorkSet Accepts a hostname, a workset name, itemlist, and
optionally a time out period. Deletes each item
in the itemlist from the workset, and if after the
deletion(s) no elements remain in the workset, delete
the workset itself. If the optional time out period is
provided, the deletion is temporary, and after the
expiration of the timeout period, the workset is
restored.
getProblemInterval Accepts a problem name and searches for the
workset entry with the corresponding problem
name to retrieve the problem checking interval.
This function usually used in the initialization
section of a rule in the expert system to get the
interval.
getWorkSet Accepts a hostname and a workset name, and
returns the itemstring containing elements
of the specified workset.
isItem Accepts a workset name and an itemstring,
examines whether the itemstring is contained in
the workset, returning TRUE if found and FALSE
if not found.
As previously mentioned, the workset program does not contain any rule declarations. 2. configs: A program which contains database declarations and routines relating to configurations. The configs program does not declare any rules. In SYSTEMWatch AI-L, configuration refers to a method of assigning threshold values and other data to a particular computer. Configurations provide a mechanism by which the system administrator can change the behavior of SYSTEMWatch AI-L without having to modify the rules of the application layer. For example, suppose SYSTEMWatch AI-L contains a rule which notifies the system administrator when the load average of a computer remains above a certain threshold so that the computer is now non-responsive. This threshold number will vary across a variety of computers because a more powerful computer can remain responsive at the same load average which might cause a less powerful computer to become non-responsive. Therefore, if a particular computer site has say 10 computers of lesser power, and 2 more powerful computers, the proper way to configure SYSTEM Watch AI-L would be to specify a lower load average for the 10 weaker computers, and a higher threshold for the 2 more powerful computers. In SYSTEMWatch AI-L configurations are specified in a text file. Thus, the system administrator can alter the threshold value used by the rules by modifying the text file containing the configuration information since after the SYSTEMWatch Al-L client has read in each of these program files, it reads the configuration file. Configurations can be specified in one of 5 formats: CONF:<hostname>:<config parameter name>:<string>:string value: Which is used to associate a string value with a config parameter name of type string. CONF:<hostname>:<config parameter name>:<num>:numeric value: Which is used to associate a numeric value with a config parameter name of numeric type. CONF:<hostname>:<config parameter name>:PROBLEM:<problem name>: Which is used to associate a configuration parameter name with a particular problem. CONF:<hostname>:<config parameter name>:SUBPROBLEM:<behavior 1>: Which is used to associate a configuration parameter name with a particular subproblem. WORK:<hostname>:<workset name>:item1:item2: . . . :itemN: Which is used to associate a workset name with a list of data. In all five formats above, the <hostname> field can either be the name of a host being managed; or it could be DEFAULT, which means all hosts except those which have a specific entry. Thus, in the example above, if the threshold for the 10 less powerful computers should be 5.2 and the threshold for the 2 more powerful computers should be 75, the following configuration declarations would be appropriate if the 2 more powerful computers had host names of server1 and server2, the config parameter name is called UNRESP LOAD AVE: CONF:DEFAULT:UNRESP LOAD AVE:NUM:5.2: CONF:server1:UNRESP LOAD AVE:NUM:7.5: CONF:server2:UNRESP LOAD AVE:NUM:7.5: The database declarations made in the configs program are, for instance, the following:
TABLE 5
ENTITY PROPERTY TYPE DESCRIPTION
CONFIG VALTYPE string The data type for a particular
configuration parameter
CONFIG STRINGVAL string The string value for a particular
configuration parameter
CONFIG NUMVAL float The numeric value for a parti-
cular configuration parameter
CONFIG PROBLEM string This value indicates the general
class of problem or type of
configuration described by this
value.
CONFIG SUBPROBLEM string This value indicates a more
specific measurement of sub-
problem as it relates to more
general configurations described
by this value.
The routines declared in the configs program are, for example, the following:
TABLE 6
NAME FUNCTION
declConfig This routine declares a configuration entry. It accepts
a host name, configuration parameter name, a value
type, a problem name, and a subproblem name.
delConfig This routine deletes from the database a particular
configuration record. It accepts a host name and a
configuration parameter name.
getConfigStr This routine returns the string value of a configura-
tion parameter name if the configuration parameter
name is of string type. It accepts a host name and a
configuration parameter name.
getConfigNum This routine returns the numeric value of a configura-
tion parameter name if the configuration parameter
name is of numeric type. It accepts a host name and a
configuration parameter name.
getConfigType This routine returns the type of a configuration para-
meter name. It accepts a host name and a configuration
parameter name.
getConfigProblem This routine returns the problem associated with a
configuration parameter name. It accepts a host name
and a configuration parameter name.
readConfigFile This routine reads a file which contains configuration
and workset declarations. It accepts a file name.
setConfig This routine sets the value of a particular configuration
parameter name. It accepts a hostname, a configuration
parameter name, and a value.
As previously mentioned, the configs program does not declare any rules. 3. events: A program which contains database declarations and routines which implements the SYSTEMWatch AI-L event handler, which allows SYSTEMWatch AI-L to execute functions either at specified times or periodically. The events program does not declare any rules. The events program defines an ordered list of records, each describing a type of event. The order is such that the next event to be executed is first on the list. Each record contains the next event time, the function to be executed at that event, and two optional values, viz., the number of instances that event is to be executed, and the interval between those instances. To add an event, an event record is added to the database. SYSTEMWatch AI-L will check for events whenever the checkEvent function is called. This function call should be placed in the main loop of the SYSTEMWatch AI-L client and the SYSTEMWatch AI-L console. The database declarations made in the events program are, for instance, the following:
TABLE 7
ENTITY PROPERTY TYPE DESCRIPTION
EVENT EVENTNAME string Unique generated name for a
scheduled event.
EVENT FUNCTION string Name of function to be executed
that the time of the event. (Func-
tion name only! - do not include
any command line arguments for
the function)
EVENT ALARMTIME integer The alarm time after which the
event gets executed
EVENT INTERVAL integer The minimum time between event
repetitions
EVENT REPEATS integer Number of times the event gets
put back onto the event queue,
after the currently scheduled
event has been executed.
The routines declared in the event program are, for example, the following:
TABLE 8
NAME FUNCTION
addEvent Given a function name, a time period, and an optional
repetition factor, addEvent schedules SYSTEMWatch
AI-L to execute the function named at a time equal to the
present time plus the time period. If the optional repeti-
tion factor is given, the function is scheduled that many
times, each time differing from the previous event time
by the time period.
checkEvent Checks the event list to see if any events are ready to
execute. If so, the ready events are executed.
delEvent Accepts a function name and removes all occurrences of
that function from the event handling system.
getNextEvent Returns the clock time to the next event waiting.
As previously mentioned, the event program does not declare any rules. 4. requests: A program which contains two routines used by SYSTEMWatch AI-L for communication between the SYSTEMWatch AI-L client and the SYSTEMWatch AI-L console. The request program only contains two function declarations. The request program provides a higher level interface to the communications system by performing some message formatting and calling the communication functions declared in the corns program discussed below. The routines declared in the requests program are, for example, the following:
TABLE 9
NAME FUNCTION
request Accepts a hostname, module name, a function name, and
arguments to the function named. Sends a message to
request the module on the host specified to execute the
named function with the specified arguments.
requestReport Accepts information which identifies a particular report
and a particular module which requested the specified
report. Formats a string containg a report request, and
sends the resulting string to the specified module.
5. coms: A program which contains routines relating to a SYSTEMWatch AI-L supplemental communications system. The corns program does not contain database declarations nor rules. The routines declared in the coms program are, for example, the following:
TABLE 10
NAME FUNCTION
manageMe Given a host name, a module name, and a
optional string, adds a SYSTEMWatch AI-L
client to the console list by calling the
addWorkSet routine. Also sends a "notifyMe"
message unless the optional string is equal to "NO
RESPONSE".
nofifyMe Accepts a hostname and a module name and adds
a process to the liveconsole list. If a process is
not on the approved console list, this function
does nothing.
sendMultiString Accepts one or more hostnames ("process list"),
a module name, a message, and a mailbox name
and sends the message muitiple times to the
modules specified on the process list
specified using the mailbox specified.
getMultiData Accepts a workset which contains a list of
processes and an entity name. This function
requests data from each of the processes on the
list of processes. The data requested is all the
data contained in the specified entity.
multiRequest Accepts a function name, parameters for the
function, and a list of at least one pair of
hostname & module name; Sends a message to
each of the hostname/module name combinations
requesting that they execute the specified function
with the specified parameters.
sendMultiManageMe Sends multiple manage me messsages to the
SYSTEMWatch AI-L consoles on the console
list.
sendMultiNotifyMe Sends multiple notify me messages to
SYSTEMWatch AI-L clients on the client list.
sendData Accepts a host name, module name, and entity
name; sendData sends all the data comprising the
specified entity to the specified host . . .
6. lib: A program which contains a series of miscellaneous routines. The libs program does not contain any database declarations nor rule declarations. The routines declared in the lib program are, for example, the following:
TABLE 11
NAME FUNCTION
fileUser Accepts a file name and returns a colon delimited list of
users which are using the specified files.
istr Accepts a floating point number and returns a string which
contains the integer portion of the floating point number.
is Accepts an optional path name. If the path name is
specified, ls returns a directory listing of the specified
path. If no path name is spec;fied, ls returns the directory
listing of the current working & rectory.
mkDirTree Accepts a directory name and creates all the necessary
directories to create the directory name specified. Thus,
if a file system only contains the root directory (/),
and mkDirTree is called with the directory name of
/A/B/C, mkDirTree creates the following directories:
/A; /A/B; and /A/B/C.
procAlive Accepts a process id and determines whether the process id
specified corresponds to a process in the process table.
readSwap Obtains the following information from the virtual memory
subsystem: swapused - the amount of swap space used on
the system. swaptotal - the total amount of swap space
allocated on the system. swapavail - the remaining amount
of swap space swapperc - the percentage of the allocated
swap is used.
systemInOut Accepts a command name and an input string. Executes
the command named using the specified input string
as the command's input. Returns a string equal to the
output of the command.
systemOut Accepts a command name, executes the command
specified, and returns a string equal to
the output of the command.
lockProcess Accepts a directory name and a filename. lockProcess
is used when you only want one process of a particular
kind to be runninhg at any one time. It guarantees process
uniqueness by first testing whether the lock file exists,
and whether it has the current process id in it. If it
has a process id in it and that process is still alive,
it returns with a warning message. If the process id
in the file is not a live process, lockProcess writes its
own process id into the file. lockProcess then re-reads the
file, and if it finds its own process id in the file,
lockProcess returns without error. Otherwise, an
error message is generated
7. alerts: A program which contains database declarations and routines used in a SYSTEMWatch AI-L alert handling system, which is used to manage problem alerts and their associated actions. The alerts program does not declare any rules. The SYSTEMWatch AI-L alert handling mechanism provides a method of presenting problem notification to the system administrator. There are several concepts used within the alert handling mechanism: 1. PRIORITY: Each alert within SYSTEMWatch AI-L has an associated priority, which is used to determine the seriousness of the detected problem. In one embodiment, SYSTEMWatch AI-L uses, for example, 5 levels of priorities comprising of FYI (least serious), NOTIFY, PROBLEM, FAILURE, and CRITICAL (most serious). 2. ESCALATION: After an alert has been created, SYSTEMWatch AI-L provides a technique of automatically changing the PRIORITY of an alert over time. This allows SYSTEMWatch AI-L to promote or demote a particular alert's PRIORITY over time. The promotion/demotion process is known as an escalation scheme. SYSTEMWatch AI-L supports the use of multiple and user-defined escalation schemes. In SYSTEMWatch AI-L, an escalation scheme is defined with a name. The escalation scheme can be associated with an alert by referencing the escalation scheme's name. SYSTEMWatch AI-L stores the escalation schemes in the configuration file. 3. TIME OUT: After an alert has been created, or after an alert has been escalated to a particular state, the technique of the present invention provides for the timing out the alert. A timed out alert is cleared from the alert system. 4. CLEAR: After an alert has been created, SYSTEMWatch AI-L provides a mechanism for clearing the alert, which removes it from the active alert pool. Cleared alerts, however, remain within SYSTEMWatch AI-L for some period of time. That period of time is called the reset time, during which if the condition which causes the alert to be generated occurs, the alert will not be posted. Once the reset time period has elapsed, the alert is completely removed from the alert handling mechanism, and if the condition which can cause the alert to be generated reoccurs, a new alert is posted. Each alert can have a different reset time. 5. IGNORE: After an alert has been created, SYSTEMWatch AI-L provides a mechanism for ignoring the alert, which, like clearing an alert, removes it from the active alert pool. Like clearing, the ignored alert is also kept within the alert handling mechanism, and has an associated time period called an ignore time, during which if the condition reoccurs, the alert will not be posted. Unlike the CLEAR mechanism, however, the IGNORE mechanism does not necessarily have a fixed ignore time for each alert. Rather, SYSTEMWatch AI-L supports an ignore scheme similar to the escalation scheme. In the ignore scheme, SYSTEMWatch AI-L remembers how many times a particular alert has been ignored. By specifying an ignore scheme, it is possible to vary the length of the ignore period depending on how many times that particular alert has already been ignored. The typical application for this is to increase the ignore period as the number of ignore actions for a particular alert has been taken. This way, SYSTEMWatch AI-L can "learn" from the actions of the system administrator and interrupt the system administrator less frequently with an alert that he has previously ignored. In SYSTEMWatch AI-L, an ignore scheme is defined with a name, and thereafter, the ignore scheme can be applied to any alert by referencing its name. SYSTEMWatch AI-L stores the ignore schemes in the configuration file. 6. ALERT ID v. ALERT REFERENCE NUMBER: Each alert in SYSTEMWatch AI-L can be identified by an alert id, which, when combined with a host name and a module name, uniquely identifies an alert, or via an alert reference number, which, when combined with a host name and a module name, uniquely identifies an alert, but only during a specific time period. In other words, the alert id is an unique number generated by SYSTEMWatch AI-L as each alert is created. On the other hand, in order for the system administrator to refer to an alert without having to type a large multi-digit number, SYSTEMWatch AI-L creates a smaller number (in one example, typically 2 digits) which points to an active alert. In order to maintain the alert reference number at 2 digits, SYSTEMWatch AI-L automatically reuses the alert reference numbers over time, so an alert reference number can only uniquely identify an alert within a certain window of time. 7. ALERT NAME and ALERT INSTANCE NAME: In addition to the alert id and the alert reference number described in the paragraph above, each alert in SYSTEMWatch AI-L can also be identified through a combination of two items, specifically the alert name and the alert instance name. The alert name identifies the class of problem which triggered the alert while the alert instance name identifies the object involved in the problem. For example, if the /usr file system reaches 90% capacity, and the fact that a file system reached 90% capacity is defined as a problem named FSFYI, then the alert name in this case is FSFYI and the alert instance name is /usr. 8. OWNER: SYSTEMWatch AI-L allows a system administrator to optionally assign owner(s) to a problem identified in an alert. This is used when the system administrator decides that someone must manually resolve the problem. Once an alert has at least one owner, the alert ceases to escalate or time out. The alert remains active within the alert handling system, and will not be removed until it is cleared. 9. PROBLEM HIERARCHIES and UNIQUE LISTS: Alerts in SYSTEMWatch AI-L may be arranged in problem hierarchies. Problem hierarchies are used to prevent a problem from triggering several overlapping alerts. For example, suppose three problems were defined as:
TABLE 12
Problem Description
FSFYI A file system reached 90% capacity
FSWARN A file system reached 95% capacity
FSALERT A file system reached 98% capacity
If a particular file system reached 98% capacity, the 3 rules which detect the FSFYI, FSWARN, and FSALERT problems would all attempt to post alert of alert type FSFYI, FSWARN, and FSALERT for the same alert instance (in this case, the file system name). However, this is redundant. What is needed is just one single alert of type FSALERT. To resolve this problem SYSTEMWatch AI-L allows problems to be grouped into hierarchies. Once a problem hierarchy has been defined, SYSTEMWatch AI-L will automatically ensure that only the alert with the most severe priority of a particular hierarchy will survive. Problem hierarchies are specified in the SYSTEMWatch AI-L configuration. In SYSTEMWatch AI-L, problem hierarchies are called unique lists. With an understanding of the above information, the operation of an alert mechanism in accordance with the principles of the present invention can now be described. When a rule detects a problem, the rule will post an alert to the alert mechanism by calling the function addAlert. During the SYSTEMWatch AI-L client's main loop, the SYSTEMWatch AI-L client will call the function checkAlert to handle alert escalation and alert clearing. When the addAlert function is called, SYSTEMWatch AI-L performs 5 validation tests before a new alert is created. In the description below, the term candidate alert refers to the alert given to addAlert for posting. The validation tests are the following: 1. Unique List Check In order to prevent a severe problem from posting related and less severe alerts, addAlert queries the database to see whether there is an existing alert with the same alert instance name and an alert name which occupies a higher priority position in the same unique list as the candidate alert. If such an alert exists, the candidate alert is rejected and not posted. 2. Duplicate Alert Check In order to prevent the problem of posting multiple identical alerts at different times, addAlert queries the database for an alert with the same alert name and alert instance name. If such an alert exists, the candidate alert is rejected and not posted. 3. Ignore List Check In order to prevent a problem of posting a new alert when the problem is being ignored, addAlert queries the database for a corresponding entry of IGNORE_IGNORETIME. If such an entry exists, addAlert compares the current clock time with the value of the entry found. An alert candidate will be rejected if the clock time is less than or equal to the value of the entry found because that condition means that the alert is being ignored at this time. 4. Clear List Check In order to prevent a problem of posting a new alert when the problem is being cleared, addAlert queries the database for an corresponding entry of ALERT_CLEARED. If such an entry exists and its value is true, then addAlert queries the database for an entry of ALERT_CLEARTIME and compares its value against the clock time. An alert candidate will be rejected if the clock time is less than or equal to the value of the entry found because that condition means that the alert was cleared and the current time is within the reset time period. 5. Lower priority check While the unique list check prevents a severe problem of also posting less severe alerts, if a more severe problem occurs after a less severe problem belonging to the same unique list has already posted an alert, the alert which corresponds to the less severe alert must be removed before the more severe alert is posted. Thus, addAlert queries the database for an alert with the same instance name and an alert name which is of a lower priority on the same unique list as the candidate alert. If such an alert is found, it is deleted. If the alert candidate passes the 5 validation tests described above, the alert will be posted. Posting an alert is a multistep process which involves the following steps: 1. Create Alert addAlert will create an alert in the database with the alert name and alert instance name corresponding to the candidate alert. 2. Add information to alert addAlert will store descriptive information into the alert. 3. Assign starting priority addAlert will query the configuration information stored in the database to retrieve the problem priority associated with an alert with the same alert name as the candidate alert and assign the same priority to the alert. 4. Assign escalation scheme addAlert will query the configuration information stored in the database to retrieve the escalation scheme associated with an alert with the same alert name as the candidate alert and store the same with the alert. 5. Assign ignore scheme addAlert will query the configuration information stored in the database to retrieve the ignore scheme associated with an alert with the same alert name as the candidate alert, and store the same with the alert. 6. Assign available actions addAlert will query the configuration information stored in the database to retrieve the available actions associated with an alert with the same alert name as the candidate alert, and store the same with the alert. 7. Assign default actions addAlert will query the configuration information stored in the database to retrieve the default actions associated with an alert with the same alert name as the candidate alert, and store the same with the alert. 8. Update SYSTEMWatch AI-L consoles addAlert will allow the alert to be communicated to the attached SYSTEMWatch AI-L consoles by calling updateNetworkAlert. 9. Save Alert to Disk addAlert will save the alert to a disk file. 10. Execute Default Action addAlert will execute any default actions associated with the alert. When the checkAlert function is called as part of the main loop of the SYSTEMWatch AI-L client and the SYSTEMWatch AI-L console, alert escalation and alert clearing are performed. Alert escalation is performed by executing the following steps for each of the alerts which has not been cleared, ignored, or assigned an owner. 1. Querying the database to retrieve the "escalation item" of an alert with the same alert name and with a priority equal to the alert's current priority. This information consists of the current priority, a time period, and a new priority. 2. If the time the alert has been in the current priority state is larger than or equal to the time period above, change the alert's priority according to the escalation scheme to the new priority. 3. If the new priority is zero, then clear the alert by removing the alert from the active alerts and place the alert on the clear list for the rest time period. 4. Determine whether any default action(s) is registered from this alert name and priority and the current time. If such a default action is registered, execute such actions by calling the doAction function. Alert clearing is performed by executing the following steps for each of the alerts: 1. Query the database for an alert to see if it has a corresponding entry of ALERT_CLEARED. If so, and if the value is true, then perform step 2. Otherwise, the process is done for this alert. 2. Query the database and retrieve a corresponding entry of ALERT_CLEARTIME. Check its value against the dock time. If that time is less than or equal to the clock time, this means the alert has been cleared, and the reset time has expired, so remove the alert. Ignoring an alert is accomplished by performing the following steps: 1. If this is the first time this alert has been ignored, store to the database a corresponding entry of IGNORE_IGNORECOUNT with value 0. 2. Query the database for an entry of IGNORE_IGNORECOUNT associated with this alert. Increment the value by one and store it back into the database. 3. Query the database for the configuration of the associated ignore scheme for this alert name and alert instance. 4. Get the Nth entry in the ignore scheme, where N is the value of the updated IGNORE_IGNORECOUNT stored in step 2 and store into the database a corresponding entry of IGNORE_NEXTTIME a value equal to the current time plus the time interval of the Nth entry. Note, if the IGNORE_IGNORECOUNT value is greater than the number of entries in the ignore scheme, put a very large number into the IGNORE_NEXTTIME. This effectively makes the ignore period infinite for all practical purposes, thereby preventing the alert from reoccurring. Note that escalation schemes and ignore schemes can be different for each managed computer by including computer specific information in the configuration database. The alerts program contains the following database declarations:
TABLE 13
ENTITY PROPERTY TYPE DESCRIPTION
ALERT PRIORITY string Describes priority of problem with
the following
words: FYI, NOTIFY, PROBLEM,
FAILURE,
CRITICAL
ALERT PROBLEMAREA string Describes the general nature of the
problem.
ALERT SHORTDESCRIPTION string Provides a brief overview of the
problem.
ALERT DETAIL string Provides a detailed overview of the
problem.
ALERT RECOMMENDFILE string Offers recommended solutions to
problem,
including useful system data.
ALERT RECOMMENDFLAG integer TRUE if RECOMMENDFILE exists.
ALERT HISTORYFILE string A cumulative problem history, save
in an outside
file. The filename is stored in this
field.
ALERT HISTORYFLAG string TRUE if HISTORYFILE exists and is a
valid file
name.
ALERT ACTIONAVAILABLE string Provides information about actions
available for
problem type. Different actions are
separated by
colons, such as 1stAction:
2ndAction.
ALERT ACTIONSTAKEN string Provides information about action in
progress and
previously taken in this alert. It
is the action
responsibility for maintaining this
field. Multiple
actions are separated by colons.
ALERT ACTIONTIME integer Contains a time stamp for when the
action should
review the current action of this
Alert. This field is
under the control of the action.
ALERT CREATTIME integer Time stamp of when the alert was
created.
ALERT CLEARED integer If a record has the cleared flag set
to TRUE, then an
alert will not be displayed as a
live alert. However,
it is still tracked in the database
to avoid immediate
realerts of the same problem.
ALERT CLEARTIME integer Time at which the cleared alert is
automatically
removed from the list and a new
problem can be
generated.
ALERT ESCALATION string Specifies name of escalation
mechanism to use for
this alert.
ALERT ESCALTIME integer Time of next escalation check.
ALERT OWNER string This is a list of people who claim
ownership for the
problem and are thereby
acknowledging the
problem's existence, which stops
problem
escalation.
ALERT PROBLEMID string Contains problem id:host:entity. For
example:
FSWARN:HOST1:/dev/sd0a. Used for
tracking if a
problem has been previously seen and
whether to
realert.
ALERT NOTIFY string Notify gets set to ADD, OWNER, or
RECOMMEND depending what value
changed.
Multiple notifications are allowed
by a colon
delimiting the notification items.
ALERT PROCESS string Specifies the owning and originating
process in the
HOST:MODULE format. The PROCESS
field with
the ALERTID uniquely specifies a
process. It is the
originating processes'
responsibility to maintain
unique ALERTIDs. By default, any
PROCESS
specified by just the HOST will
default to the
SYSTEMWatch AI-L client module.
ALERT ALERTID integer An identification number unique to
the originating
process specified in the PROCESS
property.
ALERT REFNUM integer Temporary reference number that is
used on each
local host to identify a particular
alert from the alert
displays without having to type the
whole alert
name.
IGNORE IGNORECOUNT integer Number of times the user requested
to ignore the
problem.
IGNORE NEXTTIME integer Describes the next time that
particular alert instance
may reappear if the particular
problem is noticed
again.
REFNUM REFNUM integer Contains an Alert Reference Number
allocated to a
particular local alert.
The routines declared in the alerts program are the following:
TABLE 14
NAME FUNCTION
appendAlerts Accepts a file name and saves all non-cleared alerts
to the file specified.
getAlertRefNum Accepts an alert id, assigns an alert reference
number to the alert id, and
returns that alert reference number.
getAlertHostModIdRefNum Accepts an alert reference number, and returns the
host name, module name,
and alert id associated with that alert reference
number.
checkAlertRefNum Traverses all alerts and assigns an alert reference
number to any alerts
without an associated alert reference number.
updateNetworkAlert Accepts an "old alert" and a "new alert", causes the
new alert to replace the
old alert (e.g.: delete the old alert; add new alert)
setNotify Accepts an alert, and a notification event, and sets
the alert notification flag
for the specified notification event.
doNotify Accepts an alert, checks to see whether the alert has
any notification flags set,
and if so, performs the appropriate alert
notification action.
delAlerts Accepts an alert and a scope. Scope can be local or
network wide: Deletes the
alert locally only (if scope is local), or deletes
the network both locally and
network wide.
getAlertName Accepts a host name, module name, and alert id,
returns a database key (the
name) to the alert.
clearAlert Accepts a host name, an alert id number, and a module
name; deletes the
alert specified.
ignoreAlert Accepts a host name, an alert id number, and a module
name; ignores the
alert specified.
checkProblem Accepts a problem and determines whether the problem
should be reported.
A problem will be reported if it has an alert which
is already in the database,
or if it has been specifically excluded, or if it has
been ignored.
addOwner This function is used to add an owner to a particular
alert.
getEscalTime Accepts an escalation scheme name and a current
priority, returns the
amount of time the escalation scheme specifies should
elapse before an alert
with the current priority should transition to the
next schedule priority.
getEscalTime returns this information specifically
for the host on which the
function was made if there is a configuration entry
for this host, otherwise, it
uses the escalation scheme for the "DEFAULT" host.
Returns 0 if no
escalation remains to be scheduled.
getEscalPriority Accepts a priority and an escalation scheme name,
returns the next priority
level that the specified escalation scheme will
transition to with a start state
equal to the specified priority.
getAlertNum Generates an alert identification number. The alert
identification number is
unique for a given host and module.
superceedAlert Accepts a problem name and an instance of that
problem, superceedAlert
goes through the alert mechanism and deletes any
alerts which would be
superseded by an alert with the specified problem
name and problem
instance.
addAlert This function is used to add an alert to SYSTEMWatch
AI-L.
setAlertPriority Accepts an alert and a priority, sets that specified
alert to the specified
priority.
checkAlerts Checks the alert entity for all entries which need to
be cleared or escalated.
clearAlertRefNum Accepts an alert reference number. Performs reference
number to alert
number translation, and clears the alert.
addOwnerRefNum Accepts an alert reference number and an owner.
Performs reference number
to alert number translation, and associates the
specified owner to the alert
number.
8. filesys: A program which contains database declarations, and rules used by SYSTEMWatch AI-L to monitor files and file systems on a computer. The filesys program detects, for example, the following file system problems:
TABLE 15
Problem Description Available
Actions
FSFYI File system has reached 90% full fsrecom,
rmjunk,
rmoldjunk
FSWARN File system has reached 95% full fsrecom,
rmjunk,
rmoldjunk
FSALERT File system has reached 98% full fsrecom,
rmjunk,
rmoldjunk
FSABSMIN File system has less than 1 Mb free fsrecom,
rmjunk,
rmoldjunk
FSINODEFYI File system has less than 1000 inodes free fsrecom,
rmjunk,
rmoldjunk
FSINODEWARN File system has less than 200 inodes free fsrecom,
rmjunk,
rmoldjunk.
FSINODEALERT File system has less than 20 inodes free fsrecom,
rmjunk,
rmoldjunk
FSBEHAVE1 Unusual short term behavior: File system utilization grows
by fsrecom, rmjunk,
3% in 3 minutes, as compared to the average file system
rmoldjunk
utilization for the most recent 30-minute period.
FSBEHAVE2 Unusual long term behavior: File system utilization grows
by fsrecom, rmjunk,
3% over 30 minutes, as compared to the average file system
rmoldjunk
utilization for the most recent 24-hour period.
Each of the threshold values underlined in the above table is a default value, which can be changed by the system administrator on either a computer specific basis or on a network wide basis via the configuration mechanism, as described above in the section on the config program. The FSBEHAVE1 and FSBEHAVE2 problems can only be detected if the SYSTEMWatch AI-L client can establish a historical trend line for file system space utilization. The SYSTEMWatch AI-L client performs the historical trend line evaluation by using a recursive average filter. Note: This filter can be used in areas other than file system space monitoring. Although trend line analysis can also be performed using a moving average filter, a moving average filter is less desirable than a recursive average filter because the latter can accomodate more historical data, as well as function in an environment when the sample measurement time is irregular. The recursive average filter calculates its first value by using a current value and computes subsequence instances by calculating a weighted average between the prior value and the new measurement. The weighting factor, which is called "ratio" below, may be set depending on the sensitiviy to fluctuations in the current value. The higher the ratio is set, the more the computed value will fluctuate. In SYSTEMWatch AI-L, the ratio used is dependent upon the measurement window and the time difference between the prior calculation and the current calculation. The advantage of this ratio is that it provides a filter which gives a consistent response even if the measurement intervals vary substantially. This is important, since a real time measurement system cannot necessarily guarantee accuracy in the time between calculations. One example of a recursive average filter technique is the following: Xnow=the current value of the measurement, in this case, the file system space utilization. XP=the historical value if it exists, otherwise, for the first calculation, it is equal to Xnow. Xp=(Xp.times.(1-ratio))+(Xnow.times.ratio) tdelta=current time--previous time XP was calculated ratio=1, if tdelta>time window, or tdelta/time window. The above is an embodiment of how the FSBEHAVE1 and FSBEHAVE2 rules determine the historical trend value. The FSBEHAVE1 and FSBEHAVE2 problems are detected based upon three inputs: 1. The historical trend value within a trend window; 2. The percentage difference of the current value from the trend value; and 3. The period of time over which the difference persists. A problem occurs when the difference of the current value differs by the specified amount from the trend value for a period exceeding a specified period. In the FSBEHAVE1 and FSBEHAVE2 problems, we look only at increases beyond a trend line since as far as computer file systems are concerned, drops in space utilization are not considered problems. The fileSysBehave1Compute and fileSysBehave2Compute rules function by calculating and storing a new trend value and the current time into the database. They also set a flag if the current value differs from the trend value by the specified amount. The trend values are stored in the database under the FILESYS_XP1 and FILESYS_XP2 entity/property combinations. The flags are stored in the database under the FILESYS_FL1 and FILESYS_FL2 entity/property combinations. The fileSysBehave1Test and fileSysBehave2Test will call addAlert if the time the flag was set is larger than the specified time period. The following actions are available to respond to problems detected by the filesys program:
TABLE 16
Action Description
fsrecom Analyzes a specified file system by traversing the entire file
system and gathering the following information: names of the
10 largest files, names of the 10 largest directories, the
processes using each file, the percentage of the file system
each file utilizes, names of all non-device files in the /dev
directory, names of all junk files, log files, and error files
on the file system. The files which comprises junk files, log
files, and error files are defined in the configuration. The
information gathered by the fsrecom action is stored into the
database.
rmjunk Queries database for a list of junk files producted by the
fsrecom action (see above), and removes all the junk files
retrieved from the database.
rmoldjunk Virtually the same as rmjunk (above), but only removes those
junk files whose modification time is at least 2 hours behind
the clock time when the rmoldjunk action is initiated.
The filesys program contains, for example, the following database declarations:
TABLE 17
ENTITY PROPERTY TYPE DESCRIPTION
FILESYS MOUNTPOINT string Mount point or directory name
that the file system is
mounted onto
FILESYS FSTYPE string File system type
FILESYS MOUNTED boolean Is the file system mounted?
FILESYS OPTIONS string Describes options that the file
system may be mounted with.
FILESYS SPACETOTAL integer Kilobytes of file system space
total, including space reserved
by root.
FILESYS SPACEUSED integer Kilobytes of file system space
used
FILESYS SPACEAVAIL integer Kilobytes of file system space
available to users. This number
does NOT include any in reserve
for root.
FILESYS SFACEFREE integer Kilobytes of file system space
free, including space reserved
for root.
FILESYS SPACEPERC float Percentage space used,
excluding the root reserve.
FILESYS FILEUSED integer Number of inodes/files used.
FILESYS FILESFREE integer Number of inodes/files used.
FILESYS FILESTOTAL integer Total number of inodes/files.
FILESYS FILESPERC integer Percentage of total inodes used.
FILESYS XP1 float Historical trend value for the
FSBEHAVE1 problem.
FILESYS XP2 float Historical trend value calculated
with a recursive average filter
for the FSBEHAVE2 problem.
FILESYS FL1 integer Variation flag used in the
FSBEHAVE1 problem.
FILESYS FL2 integer Variation flag used in the
FSBEHAVE2 problem.
No routines are declared in filesys. These are the rules declared in one embodiment of filesys:
TABLE 18
Rule Name Initialization Condition Then-Action
Else-Action
fileSysUpdate Sets state to DATA, gets (always true) Gathers
information on N/A
rule interval from file
systems
configuration.
fileSysAbsMin Sets state to EXCEPT, gets If file system Checks
database for file N/A
rule interval from percentages have been systems
which meet the
configuration updated since the last FSABSMIN
problem
time this rule was criteria.
For each
checked and there are problem
detected, post
file systems in the an alert to
the alert
database mechanism.
fileSysAlertFull Set state to EXCEPT, set If file system Checks
database for file N/A
ONCE to false, gets rule percentages have been systems
which meet the
interval from updated since the last FSALERT
problem
configuration time this rule was criteria.
For each
checked and there are problem
detected, post
file systems in the an alert to
the alert
database mechanism.
fileSysWarnFull Set state to EXCEPT, get If file system Checks
database for file N/A
rule interval from percentages have been systems
which meet the
configuration updated since the last FSWARN
problem
time this rule was criteria.
For each
checked, and there are problem
detected, post
file systems in the an alert to
the alert
database mechanism.
fileSysFYIFull Set state to EXCEPT, get If file system Checks
database for file N/A
rule interval from percentages have been systems
which meet the
configuration. updated since the last FSFYI
problem criteria.
time this rule was For each
problem
checked, and there are detected,
post an alert
file systems in the to the
alert mechanism.
database
fileInodeALERT Set state to EXCEPT, get If file system Checks
database for file N/A
rule interval from percentages have been systems
which meet the
configuration updated since the last
FSINODEALERT
time this rule was problem
criteria. For
checked, and there are each
problem detected,
file systems in the post an
alert to the alert
database mechanism.
fileInodeWarn Set state to EXCEPT, get If file system Checks
database for file N/A
rule interval from percentages have been systems
which meet the
configuration updated since the last
FSINODEWARN
time this rule was problem
criteria. For
checked, and there are each
problem detected,
file systems in the post an
alert to the alert
database mechanism.
fileInodeFYI Set state to EXCEPT, get If file system Checks
database for file N/A
rule interval from percentages have been systems
which meet the
configuration updated since the last
FSINODEFYI problem
time this rule was criteria.
For each
checked, and there are problem
detected, post
file systems in the an alert to
the alert
database mechanism.
fileSysBehave1Compute Set state to DATA2, get TRUE Computes
historical N/A
rule interval from trend value
using the
configuration recursive
average filter
and store
the results in
the
database under a
record of
type
FILESYS_XP1
fileSysBehave1Test Set state to EXCEP, get If the FL1 flag for a file
Adds an FSBEHAVE1 N/A
rule interval from system has been set alert.
configuration. for a time period
exceeding the
applicable time period
filsysBehave2Compute Set state to DATA2, get TRUE Computes
historical N/A
rule interval from trend value
using the
configuration. recursive
average filter
and store
the results in
the
database under a
record of
type
FILESYS_XP2
fileSysBehave2Test Set state to EXCEP, get If the FL1 flag for a file
Adds an FSBEHAVE2 N/A
rule interval from system has been set alert.
configuration. for a time period
exceeding the
applicable time period
9. files: A program which simply contains the following database declarations, which are used in tracking files and file systems.
TABLE 19
ENTITY PROPERTY TYPE DESCRIPTION
FILES DIRECTORY string Name of directory containing a file,
or the
directory name if the record
describes a directory
FILES FILENAME string Name of file or directory without
its path
FILES FILESYS string Name of file system containing file
or directory
FILES SIZE integer Size in bytes of a file or the sum
of files in a
directory
FILES LINKS integer Number of links to a file or
directory
FILES FSPERC float Percent of file system size
FILES MODE string File mode and permissions
FILES FILETYPE string File types: FILE, DIR, LINK, OTHER
FILES UID integer Owner's UID (user id number)
FILES OWNER string Owner name
FILES GID integer Owner's GID (group id number)
FILES GROUP string Group name
FILES ACCESSTIME integer file/directory access time
FILES MODTIME integer File/Directory last modificaton time
FILES PROCID string Process Ids that are accessing file
as determined by
the command fuser.
FILES PROCUSER string Process user names that are
accessing file as
determined by the command fuser
FILES PROCCOMMAND string Command name of first process on the
list
FILES DIRENTRIES string Number of directory entries in a
directory.
FILES DIRTREESIZE integer Sum of all file sizes in bytes in a
directory tree.
FILES TIMEOUT integer Time at which data should be erased
FILES COMMENT string Free form list: used primarily by
the file system
recommendation action to store class
of problem
file.
FILECHANGE DIRECTORY string Name of directory containing file,
or the directory
name if the record describes a
directory
FILECHANGE FILENAME string Name of file or directory without
its path
FILECHANGE FILESYS string Name of file system containing file
or directory
FILECHANGE SIZE integer Size of a file or the sum of files
in a directory
FILECHANGE FSPERC float Percentage of file system size
FILECHANCE MODE string file mode and permissions
FILECHANGE FILETYPE string File types FILE, DIR, LINK, OTHER
FILECHANGE UID integer Owner's UID (user id number)
FILECHANGE OWNER string Owner's name
FILECHANGE GID integer Owner's GID (group id number)
FILECHANGE GROUP string Owner's group name
FILECHANGE CREATETIME integer File/Directory create time
FILECHANGE MODTIME integer File/Directory last modification
time
FILECHANGE PROCID string Process Ids that are accessing a
file as determined
by the command fuser
FILECHANGE PROCUSER string Process Ids that are accessing a
file as determinEd
by the command fuser
FILECHANGE PROCCOMMAND string Command name of first process on the
list
FILECHANGE DIRENTRIES integer Directory entries/inodes
FILECHANGE DIRSIZE integer Sum of all file sizes in a directory
FILECHANGE DIRTREESIZE integer Sum of all file sizes in a directory
tree
FILECHANGE TIMEOUT integer Time at which data should be erased
FILECHANGE COMMENT string Free form field: used primarily by
file system
recommendation action to store class
of problem
files
FILECHANGE STARTSIZE integer File size at beginning of
measurement
FILECHANGE RATEINCREASE integer Rate of increase: (current size -
start size) /
timedelt/60.
10. swap: A program which contains database declarations, a routine declaration, and rules used by SYSTEMWatch AI-L to monitor the virtual memory swap file for problems. The swap program, for example, detects the following virtual memory problems:
TABLE 20
Problem Description Available Actions
SWAPFYI Swap space is up to 85% capacity. addswap,
tmpshutdown
SWAPWARN Swap space is up to 90% capacity. addswap,
tmpshutdown
SWAPALERT Swap space is up to 95% capacity. addswap,
tmpshutdown
SWAPABSMIN1 Available swap space is less than 5 addswap,
Mb. tmpshutdown
SWAPABSMIN2 Available swap space is less than 2 addswap,
Mb. tmpshutdown
Each of the threshold values underlined in the above table is a default value, which can be changed by the system administrator on either a computer specific basis or on a network wide basis via the configuration mechanism, as described above in the section on the config program. The following actions are available to respond to problems detected by the swap program:
TABLE 21
Action Description
addswap Increases the amount of swap space available on the
system by a two step process. First, addSwap creates
a large file by using the UNIX command mkfile. Then,
addSwap incorporates that file into the virtual memory
system by using the UNIX command swapon, which lets
the UNIX operating system to use the newly created file
as swap space. addSwap attempts to create sufficient
additional swap space so that at most 80% of the
augumented swap space is used.
tmpshutdown Shuts down the SYSTEMWatch AI-L client and console
by causing the SYSTEMWatch AI-L client and the
SYSTEMWatch AI-L console to exit their main loop
cleanswap Deletes the files added by the addswap action (above).
The following database declarations are made in swap:
TABLE 22
ENTITY PROPERTY TYPE DESCRIPTION
SWAPSTAT SWAPUSED integer Number of kb of swap space in
use. E.g.: the USED value of
the UNIX commmand pstat-s.
SWAPSTAT SWAPAVAIL integer Number of kb of swap space
available. E.g.: the
AVAILABLE value of the
UNIX command pstat-s.
SWAPSTAT SWAPPERC float Percentage of available swap
space in use. E.g.: USED/
(USED - AVAILABLE) from
the UNIX command pstat-s.
SWAPSTAT SWAPTOTAL integer Number of kb of swap space
total. E.g.: the USED +
AVAILABLE values from the
UNDIX command pstat-s.
The routines declared in the swap program are the following:
TABLE 23
NAME FUNCTION
getSwap Gathers swap space information by calling the readSwap
function, and places the information returned by the
readSwap function into the database.
These are the rules declared in filesys:
TABLE 24
Else-
Rule Name Initialization Condition Then-Action
Action
swapUpdate Sets state to DATA, gets interval (always true) Calls
the getSwap N/A
from configuration routine
swapAbsMin2 Sets state to EXCEF, gets interval if available swap is less
Posts a N/A
from configuration, gets than the limit SWAPABSMIN2
SWAPABSMIN2 limit from alert to
the alert
configuration system.
swapAbsMin1 Sets state to EXCEP, gets interval If available swap is less
Posts a N/A
from configuration, gets than the limit SWAPABSMIN1
SWAPABSMIN1 limit from alert to
the alert
configuration system
swapAlert Sets state to EXCEP, gets interval If available swap is less
Posts a SWAPALERT N/A
from configuration, gets than the limit alert to
the alert
SWAPALERT limit from system.
configuration
swapWarn Sets state to EXCEP, gets interval If available swap is less
Posts a SWAPWARN N/A
from configuration, gets than the limit alert to
the alert
SWAPWARN limit from system
configuration
swapFYI Sets state to EXCEP, gets interval If available swap is less
Posts a SWAPFYI N/A
from configuration, gets SWAPFYI than the limit alert
to the alert
limit from configuration system.
11. process: A program which contains database declarations, routines, and rules used by SYSTEMWatch AI-L to monitor processes on the computer. The process program detects, for example, the following file system problems:
TABLE 25
Problem Description Available Actions
PROCCPU1 A process is using 30% of the CPU time and the kill, stoptmp,
stopload, nice5, nice10,
system load average has reached 2.5 nice15, nice20,
schedule10, schedule25,
schedule50,
scheduleVIP10,
scheduleVIP25,
scheduleVIP50
PROCCPU2 A process is using 15% of the CPU time and the kill, stoptmp,
stopload, nice5, nice10,
system load average has reached 5.0 nice15, nice20,
schedule10, schedule25,
schedule50,
scheduleVIP10,
scheduleVIP25,
scheduleVIP50
PROCCPU3 A process is using 10% of the CPU time and the kill, stoptmp,
stopload, nice5, nice10,
system load average has reached 7.5 nice15, nice20,
schedule10, schedule25,
schedule50,
scheduleVIP10,
scheduleVIP25,
scheduleVIP50
PROCMEM1 A process is using 40% of the swap space and the kill,
stoptmp, stopload, nice5, nice10,
virtual memory system is using 80% of the available nice15,
nice20, schedule10, schedule25,
swap space. schedule50,
scheduleVIP10,
scheduelVIP25,
scheduleVIP50
PROCMEM2 A process is using 60% of the swap space and the kill,
stoptmp, stopload, nice5, nice10,
virtual memory system is using 80% of the available nice15,
nice20, schedule10, schedule25,
swap space. schedule50,
scheduleVIP10,
scheduleVIP25,
scheduleVIP50
PROCMEM3 A process is using 80% of the swap space and the kill,
stoptmp, stopload, nice5, nice10,
virtual memory system is using 80% of the available nice15,
nice20, schedule10, schedule25,
swap space. schedule50,
scheduleVIP10,
scheduleVIP25,
scheduleVIP50
Each of the threshold values underlined in the above table is a default value, which can be changed by the system administrator on either a computer specific basis or on a network-wide basis via the configuration mechanism, as described above in the section on the config program. The following actions are available to respond to problems detected by the filesys program:
TABLE 26
Action Description
kill Kills the specified process by sending the process the
UNIX kill signal
stoptmp Stops the specifled process for a specified period of time
by first sending the process a UNIX STOP signal, and
sending the process a UNIX CONTINUE signal after the
specified period of time has elapsed.
stopload Stops the specified process until the 1 minute system load
average drops beheath a specified load by first sending
the process a UNIX STOP signal, and when the system
load drops to the specified limit, by then sending the
process a UNIX CONTINUE signal.
nice5 Set the specified process' nice value to 5.
nice10 Set tbe specifled process' nice value to 10.
nicd15 Set the specified process' nice value to 15.
nice20 Set the specified process' nice value to 20.
schedule10 Reschedules a process so that it run approximately 10%
of the time. Schedule10 queries the database periodically
to ascertain what percentage of the CPU the specified
process is consuming. If the process uses more than the
goal percent CPU consumption, it is reniced such that
it uses less CPU resources. If the process uses less than
the goal percent CPU consumption, it is reniced so that
it uses more CPU resources. This action only uses
non-priviledged calls to renice.
schedule25 Similar to schedule10, except the percent CPU goal is
25% instead of 10%.
schedule50 Similar to schedule10, except the percent CPU goal is
50% instead of 10%..
scheduleVIP10 Similar to schedule10, except this action can utilize
priviledged calls to renice as well as the normal
non-priviledged calls to renice. Privledged nice calls are
those nice values which cause the UNIX operating
system to give a process more CPU time than normally
allowed. These calls are priviledged because only a
process running with an effective user id of root (the
UNIX "superuser") may assign such a nice value to
a process.
scheduleVIP25 Similar to schedule25, except this action can utilize
priviledged calls to renice as well as the normal
non-priviledged calls to renice.
scheduleVIP50 Similar to schedule50, except this action can utilize
priviledged calls to renice as well as the
normal non-priviledged calls to renice.
The process program contains the following database declarations:
TABLE 27
ENTITY PROFERTY TYPE DESCRIPTION
PROCESS UID integer User-ID of the owner of a process
PROCESS OWNER string Account name of the owner of a
processs
PROCESS MEMORY string Sum of data memory size and stack
memory size in kb,
which is all the memory usage
affecting the swap space
usage. (This does not include the
executable program
memory (text space)) See also UNIX
manual page on
the command ps(1).
PROCESS RSS integer Resident Set Size: The totaL amount of
physicaL memory
that is being used by the process at
the time, including
DATA, STACK, and TEXT. It does NOT
include
memory swapped to disk. This number
changes widely
and is under the control of the
virtual memory
manager. See aiso UNIX manual page on
the command
ps(1).
PROCESS NICE integer Process NICE priority value
PROCESS CPUPERC float Percentage of CPU used. Total CPU
used/Clock Time
since last measurement interval.
PROCESS LIFECPUPERC float Percentage of CPU Used over the life
time of the Total
CPU used / (Clock Start Time - now)
PROCESS BLOCKSIN integer Number of block input operations
PROCESS BLOCKSOUT integer Number of block output operations.
PROCESS MAJORFAULTS integer Number of major page faults. Major
page faults are
those that actually require an I/O
operation
PROCESS SYSTIME float Number of seconds of CPU Time used for
executing
system calls on behalf of this
process.
PROCESS USERTIME float Number of seconds of CPU Time used for
executing
actual user level code of a program
(i.e., the program
itself)
PROCESS TOTTIME float Number of seconds of total CPU time
used for this
process (system time and user time)
PROCESS STARTTME integer Start time of process
PROCESS COMMAND string Name of program or command executed
PROCESS COMMANDLINE string Complete command line for process
PROCCPU PROCCPU float Temporary structure used for computing
CPU % used
PROCMCT MGTYYPE string Type of management action, contains
one of: NICE,
STOPTMP, STOPLOAD, SCHEDULE.
PROCMGT RESETTIME integer Time to reset process to prior state.
Used with NICE
and STOPTMP.
PROCMGT SYSLOAD float System 1 minute load average goal.
When actual load
average goes below this value, the
process will receive
a CONTINUE signal. This is used only
with the
STOPLOAD action.
PROCMGT CPUPERC float CPU % goal to keep process. Used with
the SCHEDULE
actions.
PROCMGT OLDNICE integer Original nice value of this process,
prior to any
management actions taking place. This
is used by the
NICE actions and the SCHEDULE actions.
PROCMGT MINNICE integer Minimum allowable nice value allowed
for this
process. Used by the SCHEDULE actions.
PROCMGT MAXNICE integer Maximum allowable nice value allowed
for this
process. Used by SCHEDULE actions.
PROCMGT NICEINCR integer Increment of nice value adjustment for
this process.
Used by the SCHEDULE actions.
SYSLOAD LOAD1 float System load average over last 1 minute
SYSLOAD LOAD5 float System load average over last 5
minutes
SYSLOAD LOAD15 float System load average over last 15
minutes
The routines declared in the process program are, for example, the following:
TABLE 28
NAME FUNCTION
privNice Accepts a process id and a nice value, makes a
privileged system call to the UNIX nice function,
and reschedules the specified process with the
specified nice value.
privKill Accepts a process id and a signal name, makes a
privileged system call to the UNIX kill function,
and sends the specified process the specifled signal.
schedNice Accepts a prbcess id, a nice value, and a time
period. Checks process table to see if the process
still exists. If so, obtain its current nice value.
Then, reschedule that process with the specified nice
value. Finally store information into the PROCMGT
entity for the specified process using the following
properties: MGTTYPE shouid be "NICE"
RESETTIME should be the current time plus the
specified time period. OLDNICE should be the
specified process's nice value before we called the
nice function.
schedSTOPTME Accepts a process id and a time period. If the
specified process exists and if it is also not
currently being managed, send a stop signal to the
specified process, and store information into the
PROCMGT entity for the specified process using
the following properties: MGTTYPE should be
"STOPTMP" RESETTIME should be current
time plus the specified time period
schedSTOPLOAD Accepts a process id and a target load number. Tests
to see if specified process exists. If so, and if the
process is not already being managed, check to see
if current 1 minute load average is above the target.
If so, send a stop signal to the specified process and
store information into the PROCMGT entity for the
specified process using the following properties:
MGTTYPE should be "STOPLOAD" SYSLOAD
should be the specified load
unscheduleProc Accepts a process id. Checks to see if process still
| ||||||