Project

General

Profile

Actions

Documentation #2206

closed

Need a control command to gracefully shutdown an active MDS prior to planned maintenance.

Added by Matthew Roy about 12 years ago. Updated over 6 years ago.

Status:
Resolved
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:

0%

Tags:
Backport:
Reviewed:
Affected Versions:
Labels (FS):
Pull request ID:

Description

There is currently no way to gracefully shutdown an active MDS and allow a standby to activate or to transfer the active role to a standby (e.g. using "ceph mds"). The closest existing (documented) command is "ceph mds stop 0" which just returns "'must decrease max_mds or else MDS will immediately reactivate' (-16)".

Without a way to gracefully shutdown the active MDS or transfer away the active role, clients experience a delay after the active MDS goes down, but before a standby is promoted to the active role. This is acceptable in a failure case, but for routine maintenance there needs to be a way to reduce the delay clients experience by proactively transferring the role instead of waiting for failure to be detected.

Actions #1

Updated by Sage Weil over 11 years ago

  • Project changed from Ceph to CephFS
  • Category deleted (1)
Actions #2

Updated by Sage Weil over 11 years ago

  • Translation missing: en.field_position set to 639
Actions #3

Updated by Sage Weil about 11 years ago

  • Tracker changed from Feature to Documentation
  • Status changed from New to 12

'ceph mds fail 0'

Actions #4

Updated by Matthew Roy about 11 years ago

This is partially documented by 0c16b31db7a5ed72a9c306ae91b191c326d0776a on github.

Actions #5

Updated by Sage Weil over 6 years ago

  • Status changed from 12 to Resolved
Actions

Also available in: Atom PDF