On 26.02.2012 17:46, Sander Temme wrote:
On Feb 24, 2012, at 3:50 AM, Rainer Jung wrote:
On 24.02.2012 12:12, Steffen wrote:
Looks like the hanging "L"'s, which I reported way back, which was
investigated by Stefan
Could be but not necessarily. It depends on your MaxConnectionsPerChild setting. The
situation at w.a.o is due to processes getting recycled. I sporadically watched the
server-status there and the "L" situation did not show up for a long time. I
expect the adminds did restart Apache gracefully.
If I recall correctly, a.o httpds get a Graceful every 24h to turn over the
logs. So you should never see a really long-lived scoreboard on any of those
instances.
Yes, Daniel Shahaf also wrote on the infra list, that they have a daily
restart. But the problem seems to happen often enough, e.g. today you
can see the same symptoms. I even saw for a few seconds a server-status
that had all processes with a "no" in the "accepting" column (although I
could still get the result of the server-status itself.
Nevertheless still it is still version 2.3.15 I hope we will see
something better after infra finds time to update.
Regards,
Rainer
