| Author |
 Topic  |
|
|
jpoesen
Starting Member
5 Posts |
 Posted - 01 October 2004 : 02:48:44
|
Hi all,
I've set up three instances of the forum on the same server, in different virtual folders, each pointing to one and the same database server (different machine), but with three different forum databases ofcourse.
The problem only occurs in 1 forum.
An abnormally long delay periodically occurs in one of the forums only : when accessing a topic's post, the "T_VIEW_COUNT = T_VIEW_COUNT + 1" update statement takes about 1 minute 30 seconds, resulting in an ODBC timeout. Eventually, the statement will be executed successfully when running it from the Enterprise Manager/Query Analyzer.
The Topics table only has about 600 records, so that shouldn't be a problem. Also, I tried adding an index op TOPIC_ID, but that didn't help either.
Strange note : after I run the statement manually, everything seems to be ok for a few hours, until suddenly, the statement times out again.
I've found no locks/blocks, so I'm really puzzled.
Any idea? |
|
|
HuwR
Forum Admin
United Kingdom
20595 Posts |
Posted - 01 October 2004 : 05:06:53
|
| do you have subscriptions enabled for this forum ? |
 |
|
|
jpoesen
Starting Member
5 Posts |
Posted - 01 October 2004 : 06:11:37
|
No, we don't.
Additionaly, the same ODBC time-out now occurs when the UpdateLastHereDate() function is called in inc_func_common.asp - when trying to log in to the forum... |
|
|
|
jpoesen
Starting Member
5 Posts |
Posted - 01 October 2004 : 07:48:32
|
It seems we're having a weird problem : our transaction logs keep growing abnormally fast, while there's only *very* low forum activity.
I'm checking it out, but that's probably the problem. The DB was 6 GB, while there was only 8 MB worth of data...
I'll post my findings. |
|
|
|
HuwR
Forum Admin
United Kingdom
20595 Posts |
Posted - 01 October 2004 : 07:58:57
|
| make sure the recovery model is set to "simple" instead of "Full". |
|
|
|
Gremlin
General Help Moderator
New Zealand
7528 Posts |
Posted - 01 October 2004 : 10:45:04
|
| Enabling "Auto Shrink" may help as well .. thats an awfully large log file though, even for full logging enabled. |
Kiwihosting.Net - The Forum Hosting Specialists
|
|
|
|
HuwR
Forum Admin
United Kingdom
20595 Posts |
Posted - 01 October 2004 : 11:14:19
|
| with full it will just keep growing, it's surprising how much log space you can use up by doing a couple of bulk transfers from one db to another, but yes, 6Gb is vast. |
|
|
|
jpoesen
Starting Member
5 Posts |
Posted - 04 October 2004 : 04:14:14
|
Regarding the recovery model : it's set to "Full", but we don't want to change that. We shrunk the transaction log, but it kept growing again very rapidly (about 1 MB / minute)
Our DBA investigated some more and told me that a couple of auditing and error logging settings were enabled, so that's why everything kept growing. We've disabled those features and now the database and logfiles stay steady and the delays have not occured any more.
Thanks for sharing your thoughts, everyone.
Joeri. |
|
|
|
Gremlin
General Help Moderator
New Zealand
7528 Posts |
Posted - 04 October 2004 : 06:42:40
|
| Just curious why you woudn't change the recovery model, for an application like a forum it doesn't really add a huge amount of value imho. |
Kiwihosting.Net - The Forum Hosting Specialists
|
|
|
|
jpoesen
Starting Member
5 Posts |
Posted - 06 October 2004 : 08:09:18
|
| Nightly backups are scheduled, we'd loose a day worth of forum data if the disks crash... Why risk that if you can avoid it with a full recovery plan? |
|
|
|
Gremlin
General Help Moderator
New Zealand
7528 Posts |
Posted - 07 October 2004 : 07:02:47
|
| Of course thats a worst case scenario of 24 hours :) |
Kiwihosting.Net - The Forum Hosting Specialists
|
|
|
| |
Topic |
|
Topic Locked
