topic Re: SCSI error in Operating System - Linux

SCSI error

GnanaShekar — Sat, 08 Mar 2008 18:56:21 GMT

Hi,

We are creating linux Virtual machines on VMware ESX infrastructure.
The storage team presents LUN's to the ESX host. We create VMFS filesystem on the LUN's.
All the linux VM's we have created are using these LUN's.

One of the Linux VM seemed to be hung. We later found the /, /usr & /var file system had become read only and saw the below SCSI error message in the /var/log/messages file.

Feb 17 08:33:47 buzbrdcoll09 kernel: SCSI error : <0 0 0 0> return code = 0x20008
Feb 17 08:33:47 buzbrdcoll09 kernel: end_request: I/O error, dev sda, sector 898624
Feb 17 08:33:47 buzbrdcoll09 kernel: Buffer I/O error on device dm-0, logical block 48020
Feb 17 08:33:47 buzbrdcoll09 kernel: lost page write due to I/O error on dm-0

We informed the storage team, but they say if this was a Disk / LUN issue we should have seen this error message on all the Linux VM's that are using that LUN.

The storage team has put the ball in our court. They have asked us to find out if this issue occured due to the way Linux OS is writing data to the disk.

All the filesystems are ext3. I donot know how to proceed now.

Please suggest / point me in the right direction.

Thanks & Regards,

Re: SCSI error

skt_skt — Sun, 09 Mar 2008 00:05:36 GMT

Check if that is appliacable

All our ESX servers use multipathing for shared storage (SAN environment) and do path failover in the event of path failure. If it happens in the middle of a disk write , ESX notifies the VM's virtual SCSI controller and instruct the controller to wait. The VM interprets it as the disk is inaccessible and the disk write faults, causing an error.

There is a supported fix available online from VmWare for Red Hat Enterprise Linux 4 virtual machines. Please refer to the following link for details.

http://kb.vmware.com/selfservice/microsites/search.do?cmd=displayKC&docType=kc&externalId=51306&sliceId=2&docTypeID=DT_KB_1_1&dialogID=10922410&stateId=0%200%2010918747

Re: SCSI error

GnanaShekar — Sun, 09 Mar 2008 01:24:09 GMT

Not able to access that URL.

I get "You are not authorized to access this Document".

Find the attachment.

Re: SCSI error

GnanaShekar — Sun, 09 Mar 2008 01:31:16 GMT

Hi,

I have access to vmware site (We have platinum support from vmware for ESX and VMware Infrastructure).

Please let me know the steps to access this VMware knowledge base article.

Thanks a lot.

Re: SCSI error

skt_skt — Sun, 09 Mar 2008 02:05:47 GMT

i am not able to access too. But it was wroking when i used in Jun 2007.

It was about applying a patch to the syste. But we did not do that as it this read only scenario is very rare.I saw that only once and only one time.

Re: SCSI error

Rob Leadbeater — Sun, 09 Mar 2008 10:32:39 GMT

Hi,

Can you clarify whether you're getting this error on the physical ESX server or on the Linux virtual machine - your post appears to suggest the latter...

However it looks as though you have loaded the device multipath package, because of the references to dm-0.

You shouldn't need this package on a Linux Virtual Machine, as ESX handles all of the multipathing natively.

Hope this helps,

Regards,

Rob

Re: SCSI error

Huc_1 — Sun, 09 Mar 2008 10:39:03 GMT

Try the following link

http://kb.vmware.com/selfservice/microsites/search.do?cmd=displayKC&docType=kc&externalId=1001778&sliceId=1&docTypeID=DT_KB_1_1&dialogID=51261702&stateId=0%200%2051257565

I know nothing about your scsi issue I just started from the previous entry http link and did a little searching from there "kb.vmware.com" ... hope this get you on your way.

enjoy life.

Jean-Pierre Huc

Re: SCSI error

GnanaShekar — Mon, 10 Mar 2008 01:02:07 GMT

Hi All,
Thanks a lot for your inputs.

I have one more question?

How do I determine if there was a path failure for shared storage?

Hi Rob,

This is a Virtual machine. It is RHEL4 U3 64bit. It is allocated 2CPU & 64GB RAM.

Thanks.

Re: SCSI error

skt_skt — Mon, 10 Mar 2008 09:49:30 GMT

i am not sure if the vm guest is able to capture that. Have a look at messages file