Hello!
I'm worried about a problem that occurred twice and on different servers. We use Oracle Linux and partitions are in XFS, such a partition burst disk space and the corrupted partition. I can not mount the partition at all. Can anyone tell me why this happens?

In dmesg the error appears below.
[ 2.458820] sdf: unknown partition table
[ 2.459096] sd 2:0:5:0: [sdf] Cache data unavailable
[ 2.459099] sd 2:0:5:0: [sdf] Assuming drive cache: write through
[ 2.459262] sd 2:0:5:0: [sdf] Attached SCSI disk
[952902.649458] XFS (sdf): bad magic number
[952902.649531] XFS (sdf): SB validate failed
[953241.403987] XFS (sdf): bad magic number
[953241.404041] XFS (sdf): SB validate failed

It means your MBR is corrupted. Do you use msdos or EFI partitions?

This occurred when the disc burst space, this disk is used to store the Oracle archives.
The partition is XFS and this disk is a storage and the server is a virtual machine (vmware).

Have a look at the XFS FAQ for the discussion on Write Cache.
I don't use XFS so am not sure if this is still relevant.

Thank syg00, I'll read and post here to discover the cause of the problem.

Read the FAQ and did further research to try to understand the "write cache" could corrupt the FS, if the disk runs out of space. I found nothing enlightening.
Is the "write cache" is responsible for this bug?

Guys, it's not possible that a write cache screws up the partitions - not supposed to write outside the partitionn's boundaries.
How is the storage (partition?) mounted in the VM and what kind of VM is it?

The partition is mounted as XFS, with default options in fstab. The VM is a Oracle Linux Server 3.8.13-68.el7uek.x86_64 by vmware.

##fstab
/dev/sdf /u05 xfs defaults 0 0

But how is the partition mounted into the VM?
How do you pass (on the host) the partition to the VM?

The question is more a three part quiz. Is the hardware, filesystem or use bad?

May try a repair. http://linux.die.net/man/8/xfs_repair

XFS has been around a very long time, becoming the standard in servers maybe.

Look at this maybe for some ideas?? http://serverfault.com/questions/406...of-sparse-file

1 members found this post helpful.

pearlseattle, did not understand your question, you want to know how the disk is shown in VM?

jefro, thanks for the indication of the links, I will analyze.

What I understood is that you're forwarding a real HDD partition to a VM and that the VM is using it to store Oracle archlogs.

I'll take a step back.
The mix of what I understood and what my assumptions are is:

1. you've got a server (later referred to as "host")
2. the host has one or more HDDs
3. one of these HDDs has some partitions, and one of those is formatted with XFS
4. on the host you're running at least 1 Virtual Machine
5. that VM is running Oracle
6. you're passing the XFS partition of the HDD of the host to the VM
7. the VM mounts the partition and uses it to store Oracle archlogs
8. from time to time this specific paritition cannot be mounted in the VM

Questions:

• Is the above correct and if not, what is wrong?
• If correct, when you run the <script? whatever?> to start the VM, which parameters do you give to the hypervisor to tell it that the XFS partition of the host has to be forwarded and made available to the VM?

So far is correct.

8. It happened only once, the disk became full, thus the disk partition corrupted

But the partition is really not supposed to become corrupted if the disk becomes full... .

So just to XFS partitions? In ext partitions that does not happen ...