On Saturday, March 11, 2017 at 10:14:11 AM UTC-5, Robin Harms Oredsson wrote: > > The only pattern I've really found seems to be: > - It always happens on aborts > - It seems to be during some server load > - It only seems to happen during requests that only read data >
I don't use ZODB, but I may have an idea on where to look / possible ways to create a repeatable test-case. I think I've seen similar situations caused by transaction deadlocks or timeouts on readers high-concurrency situations when there is an extended write going on. for example, there are 50 connections reading and 1 connection writing. sometimes the writer would grab a table/row lock for too long, which would cause a pileup of locks on the readers and a few would eventually fail. under pyramid, the finished callbacks unconditionally fire after a request -- including timeouts and errors where there was never a properly setup transaction or database to cleanup. So it would try to abort, but it never got started in the first place. this may be related to what you're experiencing. -- You received this message because you are subscribed to the Google Groups "pylons-devel" group. To unsubscribe from this group and stop receiving emails from it, send an email to pylons-devel+unsubscr...@googlegroups.com. To post to this group, send email to pylons-devel@googlegroups.com. Visit this group at https://groups.google.com/group/pylons-devel. For more options, visit https://groups.google.com/d/optout.
