Project

General

Profile

Actions
Copy link

Bug #17637

closed

/usr/lib/ceph/ceph-osd-prestart.sh fails if run on real 'sh' (needs 'bash' or fix)

Added by Christoph von Stuckrad over 7 years ago. Updated over 7 years ago.

Status:
Can't reproduce
Priority:
Normal
Assignee:
Nathan Cutler
Category:
-
Target version:
-
% Done:

0%

Source:
other
Tags:
Backport:
Regression:
No
Severity:
4 - irritation
Reviewed:
Affected Versions:
ceph-qa-suite:
Pull request ID:
Crash signature (v1):
Crash signature (v2):

Description

During reboot of a ceph-server we got lots of error messages
which, at the end, only show that the access rights to journals
on disks were lost. (root:root instead of ceph:disk)

The 'correct error message (including what to do)' failed
by/at Line 55 of /usr/lib/ceph/ceph-osd-prestart.sh

if [ "${update:-1}" = "1" -o "${update:-1}" = "true" ]; then

This seems to be a 'bashism' (no precedence in old bourne shell)
Debian decided to use a real old-syntax-bourne-shell (dash), at
least during startups. So currently '#!/bin/sh' really IS 'sh',
while in older 'debian's (I remember wheezy) the default was
'/bin/bash' (even if called by 'sh').

So I think, either the script needs to be fixed or run under '#!/bin/bash'.

Stucki

Actions
Copy link
 #1

Updated by Nathan Cutler over 7 years ago

  • Status changed from New to Fix Under Review
  • Assignee set to Nathan Cutler
  • Backport set to jewel
Actions
Copy link
 #2

Updated by Nathan Cutler over 7 years ago

  • Tracker changed from Bug to Backport
  • Status changed from Fix Under Review to In Progress
  • Release set to jewel

(This bug affects jewel and hammer, but not kraken)

Actions
Copy link
 #3

Updated by Nathan Cutler over 7 years ago

  • Tracker changed from Backport to Bug
  • Status changed from In Progress to Fix Under Review
  • Start date set to 10/20/2016
  • % Done set to 0
  • Source set to other
  • Backport set to jewel
  • Regression set to No
  • Severity set to 4 - irritation
  • Release deleted (jewel)
Actions
Copy link
 #4

Updated by Nathan Cutler over 7 years ago

  • Status changed from Fix Under Review to Need More Info

Stucki: I installed dash (version 0.5.7) and the line you mentioned seems to evaluate correctly (the same as in bash) without any error.

What you describe ("During reboot of a ceph-server we got lots of error messages which, at the end, only show that the access rights to journals on disks were lost. (root:root instead of ceph:disk)") sounds like you upgraded from hammer but do not have the correct partition GUID code set on your journal partitions.

Can you post the output of "sgdisk --info=$JOURNAL_PARTITION_NUMBER $JOURNAL_DISK" for your journal partitions?

Also, can you post a log or shell excerpt showing that ceph-osd-prestart.sh really is failing at that line 55?

Lastly, I'm not sure what you mean by "The 'correct error message (including what to do)' failed".

Thanks.

Actions
Copy link
 #5

Updated by Nathan Cutler over 7 years ago

  • Status changed from Need More Info to Can't reproduce
  • Backport deleted (jewel)

No response. Feel free to re-open, providing the info requested in #17637-4

Actions
Copy link

Also available in: Atom PDF