Justin Mason wrote:
Daryl C. W. O'Shea writes:
[EMAIL PROTECTED] wrote:
Author: jm
Date: Sat May 5 05:35:44 2007
New Revision: 535514
URL: http://svn.apache.org/viewvc?view=rev&rev=535514
Log:
create temporary copies of the log files we process, to avoid race conditions
where rsyncd uploads a new rev which we then think is data from an OLD rev
Modified:
spamassassin/trunk/masses/rule-qa/corpus-hourly
This or one of the revisions up to r535536 (I haven't looked at them)
seems to have made things worse.
Sunday's active.list had a lot of rules, including all the new sandbox
rules, removed from it and (probably related) at the moment ruleqa is
only showing logs from "bb-doc jm" for r535586. I see logs from the
usual submitters for this rev on the server:
ham-bb-doc.log ham-bb-jm.log ham-daf.log ham-jm.log
spam-bb-doc.log spam-bb-jm.log spam-daf.log spam-jm.log
ham-bb-fredt.log ham-bb-zmi.log ham-dos.log ham-theo.log
spam-bb-fredt.log spam-bb-zmi.log spam-dos.log spam-theo.log
Seems to have caught up now:
http://ruleqa.spamassassin.org/20070506-r535586-n
20070506-r535586-n (Viewing)
bb-doc bb-fredt bb-jm bb-zmi cthielen daf dos jm theo zmi [+]
Was that it? was it transient (which happens if there's too little CPU
for too many logs), or are there still issues?
I guess it was transient. I would have thought that 7+ hours after all
the logs were in I wouldn't have seen this though. Perhaps if load is
this bad on the box we should consider reducing the mc-* preflight
mass-checks to 1 or 2 instead of 4. Is anyone paying attention to the
data from the mc-fast or mc-med? I found them to be to small to be useful.
I'm also still a little curious as to whether it was expected for
Sunday's active.list to drop a metric tonne of rules, and if not what
caused it to do so.
Daryl
