MySQL keep corrupting

[ekkas]

I have a problem where my SQL keep getting corrupt, then I have to rebuild the SQL tables via : Mysql Crash Recovery - Zimbra :: Wiki after which it works again for anything from a day to a few hours.

It is running
Release 6.0.9_GA_2686.RHEL5_64_20101115233514 CentOS5_64 FOSS edition.
Linux version 2.6.18-194.26.1.el5xen

On a XenServer 5.6, connected to a CentOS 5.5-64bit dedicated 'SAN' box as it's storage (for the whole VM) over 1000Mbps LAN using NFS.

I had it on local storage, but have when the problems started, I converted it from Ubuntu 8.04 32 bit to CentOS 64 bit, primarily because with the Xen kernel I can load XenServer's XenTools for better monitoring and since 32 bit is phased out, and then transferred it to SAN storage.

BUT, my problem is the SQL tables getting corrupted, the only logs I find any indication that something went wrong is in mysql_errors.log, apart from other errors that is there because the SQL tables is corrupt. Neither the XenServer nor the SAN storage have anything (that I can find) in their logs to indicate any other problems, although I'm not a Linux guru either.

I don't know if the SAN storage could be an issue, but I have a (quite loaded/busy) Windows 7 64-bit databasee 'server' VM running of the same XenServer and the same SAN/NFS without any problems.

Anyone can please help me ty to figure this animal out? I 'd really appreciate it.

Thanks

Ekkas

[phoenix]

Originally Posted by ekkas

On a XenServer 5.6, connected to a CentOS 5.5-64bit dedicated 'SAN' box as it's storage (for the whole VM) over 1000Mbps LAN using NFS.

I'm not saying this is specifically your problem but it's not recommended that you use NFS for Zimbra - take a look at this wiki article: Performance Tuning Guidelines for Large Deployments - Zimbra :: Wiki

[ekkas]

Thanks for the reply.
I had a look at that. I don't believe I have THAT big ZCS implementation. About 600 mailboxes, a lot of them unused. Ave. of about 45 GB mail a day. The Windows 7 database server I run on the same VM has a lot more IO than Zimbra. Disk read/writes and Network usage barely shows on performance graph (on XenServer) for Zimbra. So I doubt if I am stressing the system at all. I get better IO speeds on the SAN than what I get on local drive, and local drive isn't RAID'ed, so it makes me nervous to run Zimbra on a single drive.

Is there maybe something specific I can search for to find root cause of this bastardly corruption?
Do you think it might be a ghost in Zimbra somewhere (that I caused during my migration from 32 to 64 bit) or hardware related? I'd just like to find some log to narrow the options down a bit, although when I have fresh eyes, I'll look again as well.

Ekkas

[ekkas]

This morning, my Windows VM also went bonkers, so it seems my problem is not related to Zimbra.

or , I don't know.