What is the preferred way to stop services in NPM with FoE?

Our setup has two pollers with dedicated polling engine. I always try to stop the services using the Orion Service manager but as soon as I "click stop all services" each service stops for a few seconds and starts up once again on its own and returns to normal status. I suspected the FoE engine is restarting every service when it notices that the service is down and manually starts it. So I stop all services using the FoE. Is this the advised way to stop services? Why would it not obey the instruction from Service manager?

Find more posts tagged with

foe

npm

Accepted answers

All comments

FormerMember

the question is why would you want them both to stop the monitoring.

if you are patching the primary then the FOE picks up keeps the monitoring going and then when you have the primary done you patch the FoE

If you really want to shut it down thought then just go into FoE and kill it then go to the primary and shutdown to stop it.

as far as it not obeying service manager that only kills the processes on the local server so the FoE has no way to see that you are driving the shut down just that it went down so it is responding just as if the server crashed and killed the process and comes up and takes over.

Might want to open a request for them to include a function to kill the FoE in a future release

thamizh85

There are several occasions when I want to stop the monitoring. For instance if I have to do some task on my DB, I need to stop the application before taking down the DB process. I could sense the technical problem here that both FoE and Service manager seems to be de-coupled from each other and so their actions can get contradictory. But I see this as a "Use case" failure as in when I click stop services I expect the services to stop. Perhaps this would be simplified in future iterations.

cmgurley

Rather than stopping services, would an application-level "Suspend All Polling" feature address the use case? I can see this being useful in other situations, and it would keep more features within the Orion web UI. Plus, if it was an application setting, then the FoE would respect the setting since nothing would go down. It would just go quiet.

--Chris

thamizh85

That depends on whether Polling read/write is the only DB activity. If I am to shutdown the DB for maintenance, is it sufficient to suspend the polling?

On a similar note, I am curious what happens if the DB is down, would NPM stop storing all polled values or would it queue up? Say if the DB is not down, but the performance is really bad (either Low Disk I/O or WAN link congestion) how would the application server react to it? What is it's tolerance level (based on TCP timeout or read/write latency or anything else) I know this would be a different topic and I am willing to spawn a separate thread to discuss this.