Several reports that this is slow, and this is reproducible. We need to work out where the bottleneck is, but this is hard. Some possible solutions: - add more collectors - add more processors (on hadoop nodes???) - check mod_wsgi configuration - network perf between collector/processor
patch to collector to resolve the exception that aravind noted this morning is in r2506
I saw a slow submit this morning, FWIW. I watched the crashreporter client sit and spin for >30s while submitting. (I was waiting for it to quit so I could reboot Windows to apply updates.) The report says it finished submitting at 8:59 AM EDT.
Created attachment 475559 [details] at patch to correct an error in throttling The crashStorage employed by the collector included an error that would result in an exception being thrown for crashes that are not passing throttling. This corrects that problem.
Created attachment 475671 [details] [diff] [review] the addition of the procsubmitter this is the addition of the procSubmitter cron job and the rollback of collector/processor communication
we'll address the issue again at a later date