On Wed, Mar 26, 2003 at 12:54:03PM -0800, Anthere wrote:
--- Nick Reinking nick@twoevils.org wrote:
On Wed, Mar 26, 2003 at 01:53:29PM -0600, Lee Daniel Crocker wrote:
(Anthere anthere6@yahoo.com):
Say Lee, the last connexion to the french wiki
was
done at 11:21 am It was down from 11:21 to about a couple of
minutes
ago (that is 20:23) Between these two times (so ... 9 hours...), absolutely NO connexions were possible I just checked ...was the same on the en
apparently
so....which traffic ? from where ? from who ?
any idea
Well, that's the first time I'd heard that. And
there's
only one server: just because there's no traffic
on one of
the many wikis doesn't mean the server isn't busy
as hell.
The only message I got this morning was "the wiki
is down".
That didn't tell me anything, so I logged on and
saw that
the CPU load was maxed out, and that there were
hundreds
of active connections. Looked like regular traffic
to me.
I dunno Lee The wiki has been horribly slow for the past 2 weeks, and many editors have already gave up editing because of this Today, it was just as if wiki didnot existed at all during 9 hours, which were full day hours And not only was there no response from the server at all, but no response on the mailing list either. That is a bit disturbing I would say. In short, if 13/14 days it takes about 1 mn to respond for each page during day time, and the 14th day wiki doesnot exist at all, well, it won't be long for anybody to just give up
Besides, I checked on the english wiki, and I saw there were no edits in the recent changes log for many hours, so you can hardly say it is no response from "one of the numerous wiki". This was a general matter. So, we wonder.
It seems to be responding better now after the
kick, but
I'm reluctant to restart the server (which
involves breaking
all active connections) unless I get a more
specific report
than "the wiki is down".
Well, hard to expect much more than that from a user who doesn't have a login to the server itself. By the way, my guess is that some query, or perhaps some quirk, had caused MySQL to lock up. All of the apache processes were likely in a blocked state, waiting for MySQL to respond. Just because there are lots of processes running, doesn't mean they're actually responding.
-- Nick Reinking -- eschewing obfuscation since 1981 -- Minneapolis, MN
Nick, did you read that message I forwarded several hours ago from Ryo ? I was on my old email address when I did the forward, and the wikitech list refused it (the weird thing is that I just checked pending request, and there are none, so I don't know where that mail actually is:-)). But it got accepted on the main list.
In short, Ryo reported he made a query around 11 am. Just after that query, there were no answer from the server for about 5 mn. Shortly after the server finally answered, he made another one, and immediately after he made that query, the server stopped answering...for about 9 hours
these are the queries he made
select l.cur_title, r.cur_title from cur l inner join links lnk on binary l.cur_title = binary lnk.l_from inner join cur r on binary r.cur_title = binary lnk.l_to where r.cur_text like '%#redirect%'
then simply
select count( * ) from links
could this be an explanation ? or not ?
No, I didn't get the email. As far as I know, it didn't make it to the list. I'm guessing that maybe he pushed the machine beyond what it could handle, and perhaps MySQL just shat all over itself. To be fair, it is hard to expect any software to run perfectly under low resource situations. Perhaps an upgrade to a later kernel might help, but really the problem is somebody using too much of a finite resource.