Creating custom OIDs for monitoring systems on Caché using SNMP

Good day (evenings, nights, mornings - underline as necessary) to all habrogens!
This post focuses on monitoring the Caché instance using SNMP . Surely, many Caché users are already using this to some extent. Monitoring via SNMP has been supported in a standard Caché installation for a long time, but not all parameters of interest are available from the “box”. For example, I would like to see the monitoring of the number of CSP sessions, more detailed information on the use of a license, private KPIs of the system being used, etc.
In this article, you will learn how to add your own settings for monitoring Caché using SNMP.

What we already have
Caché can be monitored using SNMP. A list of what is already supported can be seen in the files <Install_dir> / SNMP directory. Usually there are 2 files: ISC-CACHE.mib and ISC-ENSEMBLE.mib . We are interested in the file for Caché - ISC-CACHE.mib . For example, we are interested in what we can get regarding licenses and sessions. The table shows the corresponding OIDs, taking into account that the hierarchy is taken from the root for Intersystems - 1.3.6.1.4.1.16563

OID	Title	Description	Data type
.1.1.1.1.10	cacheSysLicenseUsed	The current number of licenses used on this Caché instance	INTEGER
.1.1.1.1.11	cacheSysLicenseHigh	The high-water mark for licenses used on this Caché instance	INTEGER
.1.2.16	cacheLicenseExceed	A license for a license has exceeded the license available or allowed	Trap message
.1.1.1.1.6	cacheSysCurUser	Current number of users on this caché instance	INTEGER

Many important indicators are missing in the delivery, for example, the number of CSP sessions, license information is incomplete and, of course, there is no monitoring of application-specific indicators.
An example of what we would like to know:

• How many CSP users do we have?
• Limiting our license to users;
• When the license key is over.

Add a couple of parameters to evaluate the performance. The parameters themselves are in the delivery, but we want to know their increment per minute, for example:

• The increase in "global" links per minute;
• The number of commands executed per minute;
• The number of call routines per minute.

How to add "your" parameters
You can rely on the document " Monitoring Caché using SNMP ".
Our test instance (TEST) is version Caché 2013.1.3 . The operating system is Linux CentOS release 6.5 (Final) . Installing Cache on Linux OS is described in habr . The native documentation from InterSystems will also be useful.
Here is a brief work plan:

1. Create a class to collect metrics;
2. Register and activate a new class in Caché with ^% SYSMONMGR;
3. Create a custom MIB using the methods of the class MonitorTools.SNMP. To begin with, PEN (Private Enterprise Number) will take the fictitious 99990, then you need to register with IANA. This procedure is free, takes a week or two and is accompanied by a short correspondence of the form “why did you need your PEN”;
4. We start the monitoring service with the connected Caché subagent;
5. Using snmpwalk, we check that our newly created custom OIDs are available to us;
6. We add our OIDs to a third-party monitoring system. For example, take Zabbix . Zabbix documentation can be found here . By installing and configuring Zabbix specifically on CentOS, you can see it here . Make sure the monitoring works;
7. Add the launch of the system monitor in our TEST area at system startup.

And now we carry out the plan point by point.
1. Create a class to collect metrics
The metrics collection class is inherited from % Monitor.Adaptor . In the studio go to the area% SYS. Export the Monitor package. The package contains a private class Sample, which is not visible from the Studio, but is available when exporting to xml.
Suppose our work area is a TEST area. Go to her. Import the class Monitor.Sample here. We create a class in which we describe the implementation of withdrawing the 6 metrics indicated above in the section “What we already have”.

We check that the data we need is unloaded by calling the GetSample () method:

2. Register and activate a new class in Caché with ^% SYSMONMGR
Open the terminal and go to the TEST area:

3. Create a custom MIB
A custom MIB is created using the methods of the MonitorTools.SNMP class. For example, PEN (Private Enterprise Number) will take the dummy, 99990, then the PEN must be registered with IANA . View already registered rooms here . For example, PEN InterSystems has the number 16563.
16563
InterSystems
Robert Davis
rdavis & intersystems.com
To create a MIB file, we will use the MonitorTools.SNMP class, in particular, the CreateMIB () method. This method takes 10 arguments as input:

Name and type of argument	Description	What we substitute
AppName As % String	app name	The value of the APPLICATION parameter of the class metrics.snmp.Metrics - Monitoring
Namespace As % String	Our area	TEST
EntID As % Integer	PEN companies	99990 (Fiction)
AppID As % Integer	OID applications within the company	42
Company As % String	company name (uppercase)	fiction
Prefix As % String	the prefix of all the SNMP objects we create	fiction
CompanyShort As % String	short company prefix (uppercase)	fict
MIBname As % String	MIB file name	ISC-TEST
Contact As % String	contact information (in particular, address)	Leave the default value: Earth, Russia, Somewhere in the forests, Subject: ISC-TEST.mib
List As % Boolean	Analog verbose. Show the progress of the job to create a MIB file	one

Actually, the creation of a MIB file:

A new MIB ISC-TEST.mib has appeared in the <Install_dir> / mgr / TEST directory.

4. Start the monitoring service with the connected Caché subagent.
Open the System Administration Control Panel -> Security -> Services ->% Service_Monitor (click) -> Service Enabled (check) .


We also indicate that we want to start the SNMP subagent when Caché starts (click Configure Monitor Settings):

On Linux CentOS, we use the net-snmp package for SNMP monitoring. We install it, configure it on the use of subagents, and that the master agent and subagents will communicate via the standard port 705.

[root @ server ~] # grep -i agentx / etc / services
agentx 705 / tcp # AgentX
agentx 705 / udp # AgentX

A small article on the snmpd.conf configuration file in addition to the mana can be viewed on cyberciti . Here is our final setting:

Restarting Linux snmpd and snmptrapd demons. We start the ^ SNMP service to get started with the Caché SNMP subagent:
% Sys> d start ^ SNMP

5. Check that our newly created user OIDs are available .
This can be done using snmpwalk - we will show an OID that displays the number of CSP sessions:

[root @ server mgr] # snmpwalk -On -v 2c -c public localhost 1.3.6.1.4.1.99990
.1.3.6.1.4.1.99990.42.1.1.1.1.4.84.69.86.84 = INTEGER: 448035
.1.3.6.1.4.1.99990.42.1.1.1.2.4.84.69.88.8 = INTEGER: 64190
.1.3.6.1.4.1.99990.42.1.1.1.3.4.84.69.83.84 = STRING: "2014-10-11"
.1.3.6.1.4.1.99990.42.1.1.1.4.4.84.69.86.84 = INTEGER: 200
.1.3.6.1.4.1.99990.42.1.1.1.4.4.84.69.83.84 = INTEGER: 93
.1.3.6.1.4.1.99990.42.1.1.1.6.4.84.69.83.84 = INTEGER: 1

The file ISC-TEST.mib indicates the sequence of our OIDs:

FictMetricsR :: =
SEQUENCE {
fictExecutedSpeed ​​Integer32,
fictGloRefSpeed ​​Integer32,
fictKeyExpirationDate DisplayString,
fictKeyLicenseUnits Integer32,
fictRoutineLoadSpeed ​​Integer32,
fictSessions Integer32
}

Accordingly, for example, the number of sessions is the last OID 1.3.6.1.4.1.99990.42.1.1.1.6. You can compare with the number of sessions shown by SMP dashboard:


6. Add our OIDs to a third-party monitoring system.
For example, take Zabbix . Zabbix documentation can be found here . By installing and configuring Zabbix specifically on CentOS, you can see it here . Zabbix was chosen as a system that allows not only to draw graphics, but also to monitor Plain Text (in our case, this is the license expiration date and license power by users). After adding our 6 metrics to the Items of our local host and creating 4 graphs and 2 PlainText parameters (like screen elements), we see the following picture (an example of a small “live” system is shown):

At the top - information about when the license ends, and how many license slots we have. The graphs speak for themselves.

7. Add the launch of the system monitor in our TEST area at system startup
There is a good document on the use of custom routines that are triggered when Caché starts and stops. They are called respectively% ZSTART and% ZSTOP.
What we are interested in in all of this is to raise the system monitor (^% SYSMONMGR) in the TEST user area at the start. By default, this monitor starts only in the% SYS area. Accordingly, we will consider only the program ^% ZSTART. Source% ZSTART.mac (create and save it in the% SYS area).


Restart (if possible) Caché to make sure that the collection of SNMP statistics after the restart of Caché continues.
That's all. Perhaps someone will have comments on the choice of monitoring parameters or code, but the task was to show the possibility of such monitoring in principle, and you can always add the necessary parameter or refactor the code in the future.
Thanks for attention!