Daryl C. W. O'Shea writes: > On 27/03/2008 5:45 AM, Justin Mason wrote: > > Daryl C. W. O'Shea writes: > >> OK, it has got 14k+ messages, so that's OK. I do think it's a bug being > >> triggered by messages that become inaccessible or exceed the retry count > >> though. > > Yep, I missed updating the processed message count (required for the > --cs_schedule_cache version of the --server to exit) when the --server > gives up on having a message retried. > > >> On a possibly related note (it might be why we're seeing the bug being > >> triggered but not the cause of the bug), there are a small bunch of > >> these that look un-good: > >> > >> plugin: eval failed: panic: unknown regstclass 0 at > >> /home/jm/nightlymc_jm/code/masses/../rules/70_sandbox.cf, rule > >> T_FU_LONG_HTML, line 1. > > > > I (foolishly) installed perl 5.10 on the 2 slaves. looks like it's > > found a bug ;) > > Yep, and it kills the client too (and early in the mass-check), so vmsa > is doing nearly all the work.
ok. I need to file a bug, and then bring them back to 5.8.x. > Also, I noticed that the client side corpus caches are growing somewhat > large. They should probably be pruned (I usually just use 'find'). I > don't know what the expected life-span of a message is in the corpus, > though, so I'll defer this to you. what command line do you use? --j.
