[Lustre-discuss] MDT mount Problem

[Ender Güler]

Hello All,

I have lustre FS consisted of 4 servers. One of the servers is acting as mgs
and mds. This host also has OSTs mounted on itself. Role of the other 3
servers is OSS. There are no failover configuration for these OSSs and MDS.
This FS is not in production environment. So I tried to simulate some
storage based failure scienarios:

There are disk arrays  connected to OSS and MDS machine. And the OST and
MDT's are located on those disk array enclosures. I powered one of the disk
array that connected to 2nd OSS off when there was a write operation. This
made FS unreliable. So I unmounted  all the clients. There was no error of
unmounting the clients. But When I tried to stop the file system, the
MGS/MDS host and the 2nd OSS host hanged. I manually rebooted these two
hosts. Then I issued the command e2fsck on the devices that I mounted as MDT
and OSTs. e2fsck run without any fix process. But runs of e2fsck on some of
the OST devices returned the "filesystem modified" message.

After the finish of e2fsck runs, I tried to start the filesystem but the
MGS/MDS host freezed. And there are no logs about that. May be there are
some logs out there but i don't know where they are.

So could you please help me to identify what's going on? And which logs are
needed and how and where should I claim them?

Here are my environment info:

Lustre Server OS: RHEL 5.1
Lustre Version: 1.6.5.1


Thanks in advance.

Ender GULER
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lustre.org/pipermail/lustre-discuss-lustre.org/attachments/20080922/8b9884af/attachment.htm>