AS/400 Monitoring solutions

[Kos]

Simply saving the MIB file on your Zabbix server won’t be enough. The MIB just maps OIDs; the AS/400 must expose the specific data via SNMP, and you need to configure Zabbix items to query the correct OIDs. For detailed monitoring like active processes and job queue statistics, SNMP is often limited. A more reliable approach is to use scripts or APIs on the AS/400 that extract the needed data and send it to Zabbix, or run commands via SSH/Telnet and parse the results. This method gives you full control over the metrics you want to monitor.

irontmp, are you replying on the post dated by 2016 year?
Some new information was since that (this topic on the forum already has 14 pages)...

[HansB]

Hi Kos and anyone else with an idea :-)
I'm looking for a smart way to catch jobs using high CPU over a period of time.
Many systems run for long periods of time without IPL, so some "OK" jobs can have high total CPU usage (cpu seconds) time without being a problem.
What I'm trying to detect is jobs that suddenly consume a lot of CPU because of loop, bad user request (typically an SQL doing something the user didn't anticipate) etc. I would like the trigger to be able to include the exact jobname/number/user.
Has anyone created something like this that they can share?

I have looked at:
Zabbix agent metrics:
proc.cpu.util[name,user,type,subsystem,mode,jobnum]
Here - if I read the documentation correctly - I have to know each job ahead of the problem which isn't practical.

proc.cpu.util.discovery[seconds]
This will almost always after some time return the same "OK" jobs that I want to ignore.

IBM i Services:
ACTIVE_JOB_INFO (table function - https://www.ibm.com/docs/en/i/7.6.0?...table-function)
I thought of creating a discovery rule that creates an item for each job consuming over x% CPU during a specific interval. I could use RESET_STATISTICS = YES on the discovery rule, and set it to NO on the prototype items querying the CPU of a single job. I'm a bit worried that I will end up with too many items and put unnecessary load on the system.

Any other ideas?