Vultr/KVM and Centos 7 XFS corruption

Hello,

we use a standard CentOS 7.2 x64 machine, on Vultr.com Cloud environment.
We built, inside our Lab, a custom image with 2Gb drive.
When uploaded into a Cloud VPS Platform it will boot, explode and resize to fit the drive we buy.
File system is the default XFS.

With Vmware-based Host it works perfectly, never had any problems.
We are experiencing problems (one on Sunday and the other this morning), on Vultr.com that use KVM Platform.
The problems occurred on two different datacenter.

When connecting to console, that was the error:
XFS (dm-0): Internal error XFS_WANT_CORRUPTED_GOTO
Caller xfs_free_ag_extent+0x402/0x760 [XFS]
XFS (dm-0) Corruption of in-memory data detected. Shutting down filesystem...

Our kernel, before XFS corruption, was 3.10.0-327.28.2.el7.x86_64.
Now we upgraded (with standard yum) to kernel version: 3.10.0-693.5.2.el7.x86_64

Do you think it should be enough?
Are there other issues for XFS filesystem on KVM physical environment?

Thank you for everything!
F

Comments

  • Perhaps someone with more XFS experience can chime in. However, I can offer some other tips.

    The standard Vultr CentOS 7 image is built using the ext4 filesystem. Ext4 can be hot resized, but it is always more stable to resize when a filesystem is not mounted.

    A couple of suggestions:

    1. Have you tried installing on Vultr using custom ISO? For example, try installing with the CentOS 7 ISO, enabling XFS, then testing your resize procedure. That would help you rule out hardware inconsistencies between where you build your image, and launching your image on the Vultr platform.

    2. Try your xfs resize procedure while the filesystem is not mounted. This would require some additional work on your side, perhaps booting from a recovery image first.

    Best regards.
Sign In or Register to comment.