Hi,
I now actually have 2 separate customer environments where I've run into this: several processes are making connections towards http://www.terracotta.org/kit/reflector, which are then visible under requests executed in background threads. As there's no outbound connectivity to terracotta.org, the GET goes to timeout and is regarded as a failed request. When I'm sorting all the services based on failure rate, this of course messes that up when I have dozens of "Requests executed in background threads of XYZ" services with 100 % failure rate.
As a first step I tried to edit the Deep Monitoring settings:
Exclude specific incoming web request URLs
Exclude unimportant or 'noisy' incoming web requests from monitoring and alerting.
...But I couldn't get it to work; I tried several rules, but they had no effect. I suppose the magic word there is "incoming" requests - these are of course originating from the server side. Any suggestions how I could fully just exclude terracotta.org from the monitoring data? One (bad) idea is that I could manually edit the error detection settings for each and every service (there are more than 200) and classify java.net.SocketTimeoutException as an exception to be ignored. Then of course if I get "real" issues with that exception, they would also be ignored. So it's maybe not the preferred way to do it. Does anyone have any good ideas?
