Project

General

Profile

Bug #40784

mds: data may be lost when mds is restarted,and client didn't send reconnect msg during reconnect phase

Added by shen hang 4 months ago. Updated 4 months ago.

Status:
Need Review
Priority:
Normal
Assignee:
Category:
-
Target version:
Start date:
Due date:
% Done:

0%

Source:
Community (dev)
Tags:
cephfs mds
Backport:
nautilus,mimic
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(FS):
MDS
Labels (FS):
Pull request ID:
Crash signature:

Description

Assumed a client copied some obj to another location in ceph. when early_replied was received , the cp command would return normally。But the op may not be journaled when the mds was restarted. If the client didn't send reconnect msg duiring reconnect phase in time,t he obj may not exists in the destination location.
It looks like below:

:/zzz/src1
  1. cp src969 src123
    :/zzz/src1
  2. ls
    ^C^C^C^C^C^C^C^C
    :/zzz/src1
  3. ls
    src969

History

#1 Updated by Patrick Donnelly 4 months ago

  • Project changed from Ceph to fs
  • Subject changed from data may be lost when mds is restarted,and client didn't send reconnect msg during reconnect phase to mds: data may be lost when mds is restarted,and client didn't send reconnect msg during reconnect phase
  • Status changed from New to Need Review
  • Assignee set to shen hang
  • Start date deleted (07/16/2019)
  • Backport changed from nautilus,mimic,luminous to nautilus,mimic
  • Component(FS) MDS added

Also available in: Atom PDF