I'd note tho:
* the errors were massively against only *one* vhost. translations.lp.net, not bugs or answers or code
* analysing a section of the logs: of ~ 3200 5xx errors; ~ 3000 were against the translations vhost. and about 2900 of those were due solely to msnbot.
If apache is having issues it's extremely one sided; and once the bot goes away, so do the errors.
* not just 5xx errors - seeing a similar issue around 4xx's as well, just not as severe.
* when analysing the app server trace logs - the numbers from the apache logs line up:
These all had the 'A' class in the trace log if that helps as a cross verify?
eg: "A 264213136 2009-09-01T00:01:31 503 4822"
So I take this to mean that the app server is sending back the 5xx response and apache is doing it's thing with what it's been given.
yeah - it does sound similar.
I'd note tho: lp.net, not bugs or answers or code
* the errors were massively against only *one* vhost. translations.
* analysing a section of the logs: of ~ 3200 5xx errors; ~ 3000 were against the translations vhost. and about 2900 of those were due solely to msnbot.
If apache is having issues it's extremely one sided; and once the bot goes away, so do the errors.
* not just 5xx errors - seeing a similar issue around 4xx's as well, just not as severe.
* when analysing the app server trace logs - the numbers from the apache logs line up:
egrep -c "2009-09-01[^ ]+ 5[0-9][0-9] " launchpad- trace1. log.1
1063
trace2: 1090
trace3: 1016
trace4: 1080
These all had the 'A' class in the trace log if that helps as a cross verify?
eg: "A 264213136 2009-09-01T00:01:31 503 4822"
So I take this to mean that the app server is sending back the 5xx response and apache is doing it's thing with what it's been given.
???