Zimbra running slow...

[timothyalangorman]

I'm an NE user and my server is running slow..
top -u zimbra shows mysql using 2 gb and java usung around 1 gb of ram....
I have 3.5 gm of ram do I simply need more or am I seeing an issue here......
is there a cond restart for simbra?
thanks
timg

[dwmtractor]

How big of a user load are you carrying, Timothy? 3.5 GB should give you a fair amount of headroom unless you're moving quite a bit of traffic. Care to share the detailed stats of your server? There could be a number of things contributing to the slowdown.

Cheers,

Dan

[timothyalangorman]

physically it's a quad core centos 5 hp dl380
with 3.5 gb of ram...
about 110 users 40 of which are intensive users...

here's a message volume images...

[dwmtractor]

Any chance your disk array is RAID5? I've seen a number of threads about RAID5 not playing nice with Zimbra; see here. I don't know that this problem would result in high RAM usage, but I also am not certain that your high RAM usage is necessarily linked to what's going slow.

One other possibility of course, would be other hung processes, but I'm guessing since you did top you've already determined that is not the issue. Although now that I think of it, if a hung process is accessing MySQL or Java that could be it too. . .when's the last time your Zimbra services were stopped & restarted? Might try zmcontrol stop and then do another top to see if anything is still running that shouldn't. . .

Dan

[timothyalangorman]

I ended up re-booting the server. To restart zimbra nicely and kill any roue processes...
I'll need to see if I'm running raid 5 .... man I hope not....
I don't recall what I did when i staged the server.. I'll look at some notes.

[glitch23]

I have a similar problem everyday, and my only solution has been to stop zmlogger "zmloggerctl stop" and zmswatch "zmswatchctl". Once logger's mysql db is down top will start to show normal cpu usage again.

The server itself is a 2 Dual Core 2.6ghz Xeons with 4GBs of ram on a Raid-10 setup. I know the install scripts purposely set the memory limits for java and mysql so they are supposed to take up, or rather reserve as much ram as possible something like 50/50 or 50/40 with the rest left for the system.

I noticed that the logger process was sometimes running twice or more, and I believe that was causing the logger_db to spike and slow down the rest of the system.

I'm still running 4.510 until we make the jump to 5.0 maybe this was fixed in 4.5.11.