What just happened plz? was it a paperthinserver problem?

Message Bookmarked
Bookmark Removed

no sandbox where we could let the admins know there was something wrong either? :-(

StanM, Thursday, 7 June 2007 10:44 (eighteen years ago)

so it wasn't just me then. phew

Ste, Thursday, 7 June 2007 11:09 (eighteen years ago)

It is a mystery, and it is being investigated today. that's all I now right now, sorry!

Pashmina, Thursday, 7 June 2007 11:18 (eighteen years ago)

ooooooh! (ok, thx!)

http://i19.tinypic.com/632faes.gif

StanM, Thursday, 7 June 2007 11:28 (eighteen years ago)

Yes, I don't know yet. There is a sandbox, though... http://www.ilxor.com:8090/ILX/

If your firewall doesn't allow port 8090, you won't be able to see it.

Keith, Thursday, 7 June 2007 12:25 (eighteen years ago)

(if indeed I ever work it out)

Keith, Thursday, 7 June 2007 12:25 (eighteen years ago)

Well, it wasn't our server at least

  1:37pm  up 154 day(s), 11:29 

stet, Thursday, 7 June 2007 12:38 (eighteen years ago)

Yeah, I could tracert & ping it, but http timed out.

StanM, Thursday, 7 June 2007 12:46 (eighteen years ago)

I take it back, it was our problem (on the tomcat side of the church/state line :)

stet, Thursday, 7 June 2007 12:47 (eighteen years ago)

You sussed what it was? Excellent!

Pashmina, Thursday, 7 June 2007 13:25 (eighteen years ago)

Nope... Sorry.

Keith, Thursday, 7 June 2007 13:25 (eighteen years ago)

Actually, I think I've figured it now. My fault entirely. I'll fix it, stick it on the sandbox for a bit and then promote it early this evening, as it'll happen again before long, if I don't fix it.

If you're interested, the code keeps a buffer of log messages to do with statistics; how long queries take to run; how long pages take to be generated; cache hit rate and so on, but they've been next to useless, as they all just say that pages took 1ms to process and that the cache hit rate is always 99%, so I figured I would stop it writing to the database, just to save on some space (it's currently got > 200,000 rows on that table).

The buffer is 10,000 in size, and when it hits 10,000, the code does some sums to aggregate the info, and writes about 6 records to the DB. I took this code out, the other day. However, in addition to writing the data to the database, this code also clears the in memory log, which has no bounds on its size (i.e. it relies on the clear out to just not get bigger and bigger). So it basically kept on filling up until it used up all the memory and fell over.

Keith, Thursday, 7 June 2007 13:48 (eighteen years ago)

Oh right.

PJ Miller, Thursday, 7 June 2007 13:53 (eighteen years ago)

Ha ha!

Keith, Thursday, 7 June 2007 13:55 (eighteen years ago)

Well, he convinced me there!

Mark G, Thursday, 7 June 2007 15:11 (eighteen years ago)

OK, that's the fix for this up now. There are also some changes to search, and some other minor stuff. Normally, I'd like to test this stuff for longer before making it live, but under the circumstances, the fix was required to prevent it falling over again.

Keith, Thursday, 7 June 2007 16:13 (eighteen years ago)

I like the new search page. Thanks as usual.

sleeve, Thursday, 7 June 2007 16:42 (eighteen years ago)


This thread has been locked by an administrator

You must be logged in to post. Please either login here, or if you are not registered, you may register here.