Hi all, I'm using Heka to decode, filter, aggregate, and encode metrics to file, mainly using lua sandboxes: at the beginning heka works perfectly, consuming around 60% cpu cycles, but after ~40min the heka process become idle (0-1% cpu) and the following diagnostic messages starts to appear:
2015/05/18 16:20:07 Diagnostics: 38 packs have been idle more than 120 seconds. 2015/05/18 16:20:07 Diagnostics: (input) Plugin names and quantities found on idle packs: 2015/05/18 16:20:07 Diagnostics: Trwebclient1sLastAggregateMetricFilter: 38 2015/05/18 16:20:07 Diagnostics: MainRegexDispatchFilter: 32 2015/05/18 16:20:07 2015/05/18 16:20:07 Diagnostics: 25 packs have been idle more than 120 seconds. 2015/05/18 16:20:07 Diagnostics: (inject) Plugin names and quantities found on idle packs: 2015/05/18 16:20:07 Diagnostics: 3sLastAggregateMetricFilter: 23 2015/05/18 16:20:07 Diagnostics: 60sMinAggregateMetricFilter: 1 2015/05/18 16:20:07 Diagnostics: 60sLastAggregateMetricFilter: 24 2015/05/18 16:20:07 Hekad is configured with [hekad] max_message_loops = 10 maxprocs = 4 Here is a heka report after heka was started, when everything was ok: https://gist.github.com/bbinet/becc0ef3a92ad5883230 And here is a heka report after ~40min, when heka was complaining about the idle packs: https://gist.github.com/bbinet/031ea86838ae501368c4 Any idea of what is going on? Why does it work during ~40min before it starts to fail? Thanks, Bruno
_______________________________________________ Heka mailing list [email protected] https://mail.mozilla.org/listinfo/heka
