Project

General

Profile

Feature #58129

mon/FSCommands: support swapping file systems by name

Added by Patrick Donnelly 2 months ago. Updated 2 months ago.

Status:
New
Priority:
High
Assignee:
Category:
Administration/Usability
Target version:
% Done:

0%

Source:
Development
Tags:
Backport:
quincy,pacific
Reviewed:
Affected Versions:
Component(FS):
MDSMonitor
Labels (FS):
task(easy)
Pull request ID:

Description

Storage operators like Rook constantly do "reconciliation" to ensure that the desired state of the system (e.g. file system ocs-storageclass-cephfs exists) is reached. If we need to recover a broken file system (cephfs-data-scan) onto a new recovery file system, we need a way to prevent the operator like Rook from recreating the file system while we (1) rename the old damaged file system to something else for later analysis and (2) rename the recovered file system to the old name.

Right now that's not possible without introducing a new API:

ceph fs swap <fs1> <fs2>

It should operate similar to "fs rename" but operate on both file systems in one PAXOS transaction.

History

#1 Updated by Patrick Donnelly 2 months ago

  • Description updated (diff)

#2 Updated by Venky Shankar 2 months ago

The operation also needs to swap the fsid and no clients should we interfering when the swap is under execution.

#3 Updated by Patrick Donnelly 2 months ago

Venky Shankar wrote:

The operation also needs to swap the fsid and no clients should we interfering when the swap is under execution.

Hm, that's a good point. I think ceph-csi relies on the fsid.

#4 Updated by Venky Shankar 2 months ago

  • Assignee set to Rishabh Dave

Also available in: Atom PDF