[support] mysql/apache crashing

[Nat Catchpole]

I've made a bit of progress since this last e-mail.

Put back in the debian default my.cnf. mysql processes are now about 6-8% of 
memory, apache is around 3.5% per thread. However, the server is still 
swapping out, crashing etc. seemingly at random every couple of days. We're 
_only_ running drupal on it, postfix is installed but only doing sendmail 
sends from contact.module, not receiving anything, and it's otherwise a 
fairly vanilla xen/vds/debian sarge installation.

I also installed the blockcache.module to see if that would help - this has 
definitely reduced queries per page by about 20 (still roughly 200-250 
queries per page but that's 10% so not bad). There really doesn't seem to be 
any obvious trigger to the crashing out though - it'll be fine at high 
traffic levels (for us), so I'm wondering if something is going into a loop 
etc.

As always, any help much appreciated.

Nat


>
>
>I’m running only Drupal 4.7.3 on a 512mb ram VDS – debian sarge – mysql 4.1 
>apache2 php4
>
>
>
>Memory usage is generally fine overall, cpu can be a bit high sometimes 
>which I think is down to a lot of 404 errors from the old site. We want to 
>eliminate some of those 404s first, then switch off drupal’s handling of 
>404s altogether, I’m not sure this is our mysql memory problem though.
>
>
>
>Running top –u mysql  – just after /etc/init.d/mysql restart mysql 
>processes are between 4.4 and 5% of ram, then I can watch them creep 
>upwards all the way to 20% each. Eventually this slows things down, the 
>server starts swapping out, then both apache and mysql crash and require a 
>restart, then it starts over again.
>
>
>
>This may not be drupal’s fault, but I’d appreciate any advice you can offer 
>with it.
>
>
>
>Basic site information:
>
>
>
>Around 8000 nodes and 110,000 comments, private message table is around 
>10mb. Average traffic is 10 logged in users and 20 guests on line – rarely 
>more than double that, often less. Forums index takes about 300 queries due 
>to lots of subforums (c. 45) but not noticed any other pages that are 
>significant resource hogs by themselves.
>
>
>
>We’re using cck, panels, imagecache, contemplate and views a lot. This has 
>been a persistent problem for months now, which I originally put down to 
>cache table overhead (sometimes megabytes of overhead within a few minutes 
>of flushing, have dbmaintenance optimising it every hour now).
>
>
>
>I’m using a fair amount of contrib modules, so am wondering if it could be 
>down to one of those notwithstanding my mysql settings. Everything is 
>pretty much up-to-date by the way – using almost entirely 4.7 versions 
>except for a couple running CVS where there’s not an official 4.7 branch.
>
>
>
>Full list of contrib. modules below, devel module is generally off. Also 
>making use of free tagging, forums, tracker and other core modules. 
>Forum_get_forums runs pretty slow judging by devel module’s output, and 
>administering taxonomy access requires a php memory limit of around 60mb 
>due to free tagging, so I’m wondering if it’s something to do with one of 
>those.
>
>
>
>bbcode
>
>tagadelic
>
>cck
>
>flatforum
>
>pathauto
>
>taxonomy_access
>
>comment_mover
>
>image
>
>privatemsg
>
>contemplate
>
>imagecache
>
>quicktags
>
>views
>
>content_moderator
>
>imagefield
>
>quote
>
>db_maintenance
>
>nodewords
>
>scheduler
>
>devel
>
>panels
>
>smileys
>
>
>
>For completeness, here’s my my.cnf settings:
>
>
>
>skip-locking
>
>key_buffer_size = 20M
>
>max_allowed_packet = 1M
>
>table_cache = 384
>
>sort_buffer_size = 1M
>
>net_buffer_length = 8K
>
>read_buffer_size = 256K
>
>read_rnd_buffer_size = 1M
>
>myisam_sort_buffer_size = 4M
>
>query_cache_type = 1
>
>query_cache_size = 64M
>
>query_cache_limit = 4M
>
>max_connections = 55
>
>max_user_connections = 55
>
>
>
>Any help much appreciated. This problem started happening when I had very 
>few modules installed (mainly flatforum/smileys/cck/views/taxonomy 
>access/privmsg/quicktags/bbcode).
>
>
>
>Thanks!
>
>
>
>Nathaniel
>
>