Bug 1020297 - Reduction of 2 legged mirror while the 3rd. leg is being synchronized breaks the mirror.
Reduction of 2 legged mirror while the 3rd. leg is being synchronized breaks ...
Product: Fedora
Classification: Fedora
Component: lvm2 (Show other bugs)
Unspecified Unspecified
unspecified Severity unspecified
: ---
: ---
Assigned To: Jonathan Earl Brassow
Fedora Extras Quality Assurance
Depends On:
  Show dependency treegraph
Reported: 2013-10-17 08:06 EDT by Zdenek Kabelac
Modified: 2016-07-19 06:29 EDT (History)
11 users (show)

See Also:
Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of:
Last Closed: 2016-07-19 06:29:52 EDT
Type: Bug
Regression: ---
Mount Type: ---
Documentation: ---
Verified Versions:
Category: ---
oVirt Team: ---
RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: ---

Attachments (Terms of Use)

  None (edit)
Description Zdenek Kabelac 2013-10-17 08:06:28 EDT
Description of problem:

When lvconvert -m-1 is executed on mirror which is not yet fully synchronized,
log removal cause abort of conversion process.

This can be observed in lvm2 internal test suite
(make check_cluster T=lvconvert-mirror.sh):

lvconvert-mirror.sh: line ~190  

lvconvert -m-1 $vg/$lv1 "$dev2"

If this conversion is execute just in time, convertion from 
the command above -m+1 is not yet fully synchronize, it
will fail together with signs, there are still some mimagetmp volumes:

metadata/lv_manip.c:696   Stack LV1:0[0] on LV LV1_mimagetmp_2:0
metadata/lv_manip.c:300   Adding LV1:0 as an user of LV1_mimagetmp_2
metadata/lv_manip.c:696   Stack LV1:0[1] on LV LV1_mimage_2:0
metadata/lv_manip.c:300   Adding LV1:0 as an user of LV1_mimage_2
metadata/lv_manip.c:696   Stack LV1_mimagetmp_2:0[0] on LV LV1_mimage_0:0
metadata/lv_manip.c:300   Adding LV1_mimagetmp_2:0 as an user of LV1_mimage_0
metadata/lv_manip.c:300   Adding LV1_mimagetmp_2:0 as an user of LV1_mlog
locking/cluster_locking.c:428   Requesting backup of VG metadata for @PREFIX@vg
format_text/archiver.c:226   WARNING: This metadata update is NOT backed up
metadata/mirror.c:1102   Failed to initialize log device

Version-Release number of selected component (if applicable):

How reproducible:

Steps to Reproduce:
1. lvconvert -m+1 
2. right after that  lvconvert -m-1 while the first conversion is not fully done.

Actual results:

Expected results:

Additional info:
Comment 1 Zdenek Kabelac 2013-10-18 05:34:34 EDT
So the problem looks to be related to how the mirror evaluates it's still a mirror.

If the mirror has just 2 legs - and I add 1 leg - which is slowly being synchronized, and while the synchronization is still in progress, I cut other the one of those in-sync legs  - the mirror tries to demolish itself.
This is catched by clvmd test which is missing the lock from log_lv,
thus  lvconvert attemt to deactivate this LV (which has no lock) fails -

that's probably the reason for the weird sequence in the  lvm2 code:
(lib/metadata/mirror.c - line ~300)

	/* If the LV is active, deactivate it first. */
	if (lv_is_active(log_lv)) {
		(void) deactivate_lv(cmd, log_lv);
		 * FIXME: workaround to fail early
		 * Ensure that log is really deactivated because deactivate_lv
		 * on cluster do not fail if there is log_lv with different UUID.
		if (lv_is_active(log_lv)) {
			log_error("Aborting. Unable to deactivate mirror log.");
			goto revert_new_lv;

Which is fails in cluster since without lock no deactivation is called - so lv will remain active and the second test fails.

This is sign of multiple bugs here - mirror needs to be deactivated through it master LV - and it should never go to the subdevices.

And also we have here the problem with detection what is and what is not a mirror. So if the mirror is being upscaled, and the removal of the leg would cause device to be no longer mirrored - seems we have multiple choices -
let it go -  mirror will catchup later to be a mirror?
abort instantly - and warn user he needs to wait till mirror has synced its legs?

Here is short 'test-suite' code to demonstrate the problem:

. lib/test

aux prepare_pvs 4
vgcreate -s 32k $vg $(cat DEVICES)

# "remove from original mirror (the original is still mirror)"
lvcreate -aey -l10 --type mirror -m1 -n $lv1 $vg "$dev1" "$dev2" "$dev3"
aux delay_dev "$dev4" 0 100
lvconvert -m+1 -b $vg/$lv1 "$dev4"
lvs -a -o+devices $vg

lvconvert -m-1 $vg/$lv1 "$dev2"
lvs -a -o+devices $vg

vgremove -ff $vg
Comment 2 Jaroslav Reznik 2015-03-03 10:08:51 EST
This bug appears to have been reported against 'rawhide' during the Fedora 22 development cycle.
Changing version to '22'.

More information and reason for this action is here:
Comment 3 Fedora End Of Life 2016-07-19 06:29:52 EDT
Fedora 22 changed to end-of-life (EOL) status on 2016-07-19. Fedora 22 is
no longer maintained, which means that it will not receive any further
security or bug fix updates. As a result we are closing this bug.

If you can reproduce this bug against a currently maintained version of
Fedora please feel free to reopen this bug against that version. If you
are unable to reopen this bug, please file a new report against the
current release. If you experience problems, please add a comment to this

Thank you for reporting this bug and we are sorry it could not be fixed.

Note You need to log in before you can comment on or make changes to this bug.