MySQL Server crashing more and more often as site grows

[findingpeace]

On pretty much a biweekly basis now, our vB4 site is crashing and I get a million "vBulletin Database Error!" emails. Most of the time, the database gets corrupted and needs to be repaired. I'm providing my server & forum stats below, in hopes that someone can recommend some tweaks for me to stop all of this downtime?


FORUM INFO
vBulletin Version: 4.2.2 PL1
Total Size: Threads 17,291 Posts 247,138 Members 7,945
Posts per day: 800
Google Analytics Daily Visitors: 5,000
Google Analytics Daily Page Views: 30,000
Resource Intensive Mods: DBTech vBNotifications (15 second refresh)


SERVER INFO
KnownHost VPS SSD-3
3584MB Guaranteed RAM
50GB RAID10 SSD Storage
3x Priority 24 CPU Cores
4500GB Premium Bandwidth

MySQL conf file:
thread_cache_size=4
table_cache=512
local-infile=0
innodb_file_per_table=1
query_cache_limit=1M
max_connections=500
key_buffer_size=16M
join_buffer_size=1M
query_cache_type=1
query_cache_size=32M
max_tmp_tables=1
default-storage-engine=MyISAM
open_files_limit=4974
max_allowed_packet=268435456
innodb_use_native_aio=0
innodb=on
innodb_buffer_pool_size=134217728


People automatically say "raise the refresh rate of vBNotifications!" but this really should not be a problem. Our site, even at peak hours, runs at only 1-4 CPU load average (10-20% total), and 1GB memory free.

The site just randomly craps out and crashes. Nothing to do with unusually high load.

Any thoughts?
Thanks!

[ForceHSS]

Have you talked to your host yet about this as they might be able to fix the problem

[findingpeace]

Have you talked to your host yet about this as they might be able to fix the problem

Yep, they blamed it on the site :D

[ForceHSS]

Do you have any server error logs from your host and did your host even tell you what was causing it as many hosts always blame it on the software not themselves

[findingpeace]

Do you have any server error logs from your host and did your host even tell you what was causing it as many hosts always blame it on the software not themselves

They said it was due to a spike in memory usage. Not totally sure what most of this means:

The first OOM occurrence today was at 'Aug 14 10:26:18'

--
Aug 14 10:26:18 host kernel: [8070457.454566] Out of memory in UB 11679: OOM killed process 5283 (mysqld) score 0 vm:522564kB, rss:269476kB, swap:0kB

The first occurrence of the database being corrupt was at '140814 14:47:39'

--
140814 14:47:39 [ERROR] /usr/sbin/mysqld: Incorrect key file for table './mysite/vbhumanverify.MYI'; try to repair it

The daily process log in WHM -> Server Status -> Daily Process Log, shows apache (the webserver) which runs as the user 'nobody' using the most RAM (44.45%).

This further points to the servers web activity which is the most likely source of the memory being consumed.

If you continue to experience out of memory issues, I would recommend upgrading your VPS package to allow for the RAM your processes require.

[ForceHSS]

You will need to get your host to fix these problems and get them to give you more memory. I am sure there are others on here that can give you more detail info on these problems

[Zachery]

Should really go post on .com for a server optimization.