we have an ESP v11.4 appl consisting of about 60 AGENT_MONITOR types to alert us when a cybAgent is not communicating. On weekends our unix team does patching and a select number of boxes (and thus agents) are down for a few hours...different boxes on different weekends.
Is there a way put a HOLD on an AGENT_MONITOR ahead of time for those patching hours so we don't page out?
The only way I'm aware of to throttle an agent_monitor is to post it complete in the non-stop running appl but then we'd have to complete and restart the entire appl in order to get those individual agent_monitors listening again.
here is an example of one of the 60
AGENT_MONITOR NT_YADA STATINTV 5 MSGQLEN 10 AGENT NT_YADA AMNOTIFY ALERT(AGNT) RUN ANYENDJOB
You can issue "ESP OPER ESPCOM DEST(NT_YADA) QUIESCE" in the alert related event. It will stop sending messages to the agent.
After the agent is restarted, issue OPER ESPCOM DEST(NT_YADA) RESTART from ESP page mode to resume.
Hope this helps.
Happy New Year!
Hey Lucy, Happy New Year. So no way to put an AGENT_MONITOR type on hold or anything?
I’m splitting hairs here because, I’d have much less of a compunction to put one, lone agent_monitor job on hold, where I could simply release it, say the next morning, as opposed to quiescing all traffic to an agent that I’d have to possibly hang around till the wee hours to restart it. (naturally, if we didn’t quiesce the queued up jobs would take off on their own without intervention…not so if I issue the quiesce; Then you know who would likely have to hang around to issue the restart command.)
I’m guessing agent_monitor types don’t adhere to some of the csf commands like normal jobs, but had to ask anyway.
Unfortunately the HOLD command won't work for agent_mon, which will run every 5 minutes (as specified on STATINTV) automatically after it runs.
I would think it's a better idea to quiesce the agent, otherwise the queued up jobs may fill up the COMMQ file.
If you just don't want to be paged, then maybe you can suspend the related event.
Hope this helps,
I agree with Lucy that holding all work going to the agent is a better solution. As she mentioned it is possible the COMMQ could get filled up if there were significant amount of work going to this agent.
Just a different thought on the HOLD/RELEASE
It would be possible to force complete the agent monitor and then restart the agent monitor when the outage is complete.
Do you have any more concerns on this question?
Thank you for your feedback,
Another option is to modify the ALERT(AGNT) routine to not notify (maybe not page but send an email) during the known maintenance window.