I have a problem with a VM that keeps going into read-only mode on Fedora 4:
Linux 2.6.17-1.2142_FC4 #1 Tue Jul 11 22:41:14 EDT 2006 i686 i686 i386 GNU/Linux
... I have checked out a few links, but on this kernel level, it's not helping... the links are:
Anyways, what I am getting is journal aborts, and other similar errors, then the file system goes into read only. Any help would be much appreciated!!!!
Check /var/log/vmkernel for SCSI errors. If your storage is having issues the filesystems in Linux VMs go read-only. For example, when my SAN has issues, my Linux VMs go read-only.
Edward -- What kind of vmkernel issues where you seeing? We have an MSCS cluster that was giving us problems, and I wonder if you were seeing the same thing - with scsi aborts/resets on the SAN...
More importantly, I was hoping to find a patch...
There's a problem with the scsi driver and it going into panic mode when it doesn't get a response from the virtual disk. The result is it goes read-only to protect the filesystem.
The above mentioned KB has a patched driver to alleviate this behavior.
would downgrading the kernel be an option then? It all boils down to the behavior of the SCSI driver.
edit to clarify; try grabbing the kernel from a centOS install disk; install that; then the patch from the KB.
Message was edited by:
I have experienced read-only file system with CentOS 4.4 version too. Also with ESX v3.0.2 host that has RHEL5 running as a guest have similar problem. So it seems that the problem is bigger than stated in KB article 51306 ( http://kb.vmware.com/kb/51306 ) ...
If the Storage becomes unavailable (SCSI Reservation Conflict) or storage dissappears then the filesystem goes read only to protect the filesystem.
The only time I see this is when the storage has issues or there are LOTS of SCSI reservation conflicts.
You might say that I figured out an interim solution - at best.
Since I had a Microsoft Cluster on the same VMware ESX cluster as my Fedora VM, I decided on my SAN to create a new host group (Read: LUN Masking) for just my ESX/MSCS cluter nodes.
This solved the issue, but I won't have that much luck on another ESX cluster that I manage ... sigh