Server is: Debian 6 / Xen / 2GB RAM / BOA Latest Stable Upgrade from 2.0.1 > 2.0.2

I updated one of our aegir servers today and it is now chewing up every bit of RAM available on a regular basis. If the usage backs off it is only short lived then it jumps back up, this has been constant for over 24 hours now.

On servers with 2GB that used to run at 50% most of the time, now run at 90% most of the time. There is less traffic on the sites today than there was yesterday so a wee bit confused as to how to report the issue. Any suggestions would be very helpful.

Comments

omega8cc’s picture

omega8cc commented

There is nothing we could suggest if you don't provide any output from top or htop command.

Try to disable Redis/Memcached completely:

1. First add $use_cache = 'NO'; line to your /data/conf/override.global.inc file.
2. Then run bash /var/xdrago/clear.sh twice.
3. Now check the RAM usage.

Also, to find which site(s) is causing this, type as root: telnet localhost 6379, then monitor and watch for the flood of cache requests.

snlnz’s picture

snlnz commented

Thanks for the reply but I thought I'd wait for your response before pasting endless information into the queue. :)

I disabled Redis/Memcached as suggested, then ran clear.sh twice.
Ram usage dropped to below 10% immediately, awesome! So what do you suggest from here?
Should I re-enable Redis/Memcached or is there something else I need to do first?

I also telnet localhost 6379 and typed monitor, but no cache requests at all.

omega8cc’s picture

omega8cc commented

When you disabled Redis for all sites, there will be no requests in the Redis monitor, of course. You may want to enable it again if you wish to see anything there, and then you will see which site exactly causes the issues. Then disable it only for this site.

omega8cc’s picture

omega8cc commented
Status: Active » Fixed

We no longer use the old version of Redis and the caching backend is now simplified and uses only latest Redis version.

Status: Fixed » Closed (fixed)

Automatically closed -- issue fixed for 2 weeks with no activity.