I updated to 10.3.9, and no problems for 36 hours now.
When it hits 48 hours, I'm going to take my router blocks out (I had
found that blocking certain spiders seemed to improve the situation)
and closely monitor it for another couple days.
In the interest of completeness:
From: Niels Meersschaert <email@hidden>
Have you tried eliminating the dynamic sites (php &
mediawiki) & see if the problem continues?
I'll have customers out of action long
enough to see what is happening. Considering I can go a whole day
without a problem, that's unacceptable at the moment.
Did you start to notice the behavior after you
added the dynamic sites?
No. My gallery system (PHP/MySQL) has been running for two years now.
MediaWiki has been running since January.
Can you
configure the three sites into different log files so that you can
isolate requests?
Yes, I set up all the sites I host to have separate log files. When
the load shoots up, I was doing:
ls -t /Library/WebServer/*/logs/access_log
to see which one had been hit recently, then did a "tail" on that log
file.
(As near as I can tell, the Server only
buys me a fancy admin GUI for what I'm doing.)
Actually the server version does give you more than a fancy GUI over
the client. The OS itself is designed for operation as a server &
thus has different tuning for memory, network, services, etc that
provides better throughput than you get on client as a server.
Although I didn't pay $500 for OS Server, this is a dedicated
internet server machine, and I've done a fair amount of manual
tuning. It normally does a great job.
your machine might be underpowered.
G4/350 was fine until I installed MediaWiki... :-(
If you can disable MediaWiki & see if the problem returns, it will
help isolate the issue.
Sorry, I did not mean to imply that MediaWiki was involved. I've been
running it since January, and "the problem" only showed up in the
past couple weeks.
MediaWiki is resource-intensive in general, though. I do want to get
memcached installed someday.
netinfo.log has a lot of:
Jun 14 04:28:57 dns lookupd[5986]: NetInfo connection failed for
server 127.0.0.1/local
Jun 14 04:37:24 dns last message repeated 6 times
...
This section of the log implies that a process is attempting to
connect to a server at 127.0.0.1 for NetInfo data & is failing.
Every OS X machine runs a local NetInfo server(localhost=127.0.0.1),
so this seems to say that it can't connect to it. Any chance NetInfo
was touched?
No, it was not.
If not, do you get these messages after reboot, or only
when lookupd hogs CPU?
Only when the problem manifests. I had re-booted several times, which
got rid of the problem only temporarily, and the logs do not indicate
the NetInfo failure immediately.
Since you said you were running the DNS server on this machine, could
be that your DNS isn't fully configured. OS X likes to have fully
qualified DNS in both directions. Thus you need not only an A
record, but also the reverse for your IP.
That all seems to be working fine. I've been running DNS for a decade
(on Solaris prior to MacOS X), and I do think I have it setup properly.
Is the BIND server authoritative for it's own zone & caching
other zones...
Yes.
What DNS
server is it using for lookups outside?
I'm using my ISP's nameserver. I believe it's NS1.SPEAKEASY.NET.
Are there performance issues
on that server? Can you try another server?
Whenever I have DNS-related issues, the FIRST thing I do is nslookup
with some other server! :-)
Is the machine setup to
use itself as it's DNS server for lookups?
Yes.
Try networksetup to check
that.
That must be part of OS X Server -- no "networksetup" on my machine!
I added MediaWiki in January, and it's run smoothly (well yea, a
bit slow) until just a couple weeks ago.
Aside from MediaWiki, was there any other change to the machine prior
to a few weeks ago? Was there a recent spike in traffic? Was the
machine rebooted?
The ONLY thing that really has changed is that I added a couple more
zones, and asked my ISP to back them up. But that was not coincident
-- it was 4-5 weeks ago, and the problem just stared a couple weeks
ago. But it was roughly coincident with when my ISP started doing
zone transfers on the new domains -- but I chalk that up to
coincidence, since they've been doing periodic zone transfers on my
other domains for years now.
I guess I should also stop and install 10.3.9.
10.3.9 did add some network fixes, so it's probably a good idea to
put that in place.
And it seems to have fixed it... knock wood!
I guess I'll never know what happened, which troubles me, but I'm not
nearly as troubled as I would have been if 10.3.9 didn't seem to do
the trick!
Thanks again for all the great ideas and help! Even though none of
them turned out to be the "smoking gun", they were all helpful in
focusing my attention on the problem.
:::: Those right-wing, religious nuts have gone too far: in Texas, a
man can't even marry his widow's sister!
:::: Jan Steinman <http://www.Bytesmiths.com/Van>
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Web-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
http://lists.apple.com/mailman/options/web-dev/email@hidden
