Project

General

Profile

Actions
Copy link

Bug #9826

closed

ceph osd crush rule ls should use the pending crush, if any

Added by Loïc Dachary over 9 years ago. Updated over 9 years ago.

Status:
Rejected
Priority:
Normal
Assignee:
Loïc Dachary
Category:
Monitor
Target version:
-
% Done:

80%

Source:
other
Tags:
Backport:
Regression:
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

The following is racy:

ceph osd crush rule create-simple ...
ceph osd crush rule ls

because osd crush rule ls does not use the pending crushmap. If the
machine is fast enough the second command will run before paxos has time
to propose and incorrectly claim the rule does not exist.

Related issues 1 (0 open — 1 closed)

Related to Ceph - Bug #9858: osd crush rule create-erasure idempotency failureRejectedLoïc Dachary10/21/2014

Actions
Actions
Copy link
 #1

Updated by Loïc Dachary over 9 years ago

  • Description updated (diff)
Actions
Copy link
 #2

Updated by Loïc Dachary over 9 years ago

  • % Done changed from 0 to 80
Actions
Copy link
 #3

Updated by Loïc Dachary over 9 years ago

  • Status changed from In Progress to Fix Under Review
Actions
Copy link
 #4

Updated by Loïc Dachary over 9 years ago 

<sage> loicd: 9826, i think the current behavior is correct
<sage> the first command won't return until the paxos round has committed
<sage> so the ls will always see the result.
<sage> in general, we don't want to ever expose uncommitted state to clients (because the mons may fail/restart before it commits)
Actions
Copy link
 #5

Updated by Loïc Dachary over 9 years ago

  • Status changed from Fix Under Review to Rejected
Actions
Copy link

Also available in: Atom PDF