Back to bug 1009700

Who When What Removed Added
Corey Marthaler 2013-09-27 16:27:44 UTC Summary killing both redundant logs causes deadlock (take 2) mirror device failure scenarios lead to deadlocks
Corey Marthaler 2013-09-27 16:42:15 UTC Keywords TestBlocker
Priority unspecified high
Alasdair Kergon 2013-10-01 21:47:11 UTC Assignee lvm-team jbrassow
Jonathan Earl Brassow 2013-10-01 21:50:22 UTC Status NEW ASSIGNED
Tom Lavigne 2013-10-22 12:46:18 UTC CC tlavigne
Peter Rajnoha 2013-10-23 15:27:59 UTC Status ASSIGNED MODIFIED
Fixed In Version lvm2-2.02.100-7.el6
errata-xmlrpc 2013-10-23 15:40:19 UTC Status MODIFIED ON_QA
Jonathan Earl Brassow 2013-10-23 22:57:52 UTC Doc Text A small window of opportunity exists for a bug to express itself during device failures while using LVM mirrors. It is possible for LVM commands to become hung when attempting to read an LVM mirror just after a write failure to a device but before the repair command has had a chance to handle the fault.

A new configuration option has been made available to avoid this issue - the 'ignore_lvm_mirrors' option in the lvm configuration file. Setting this to '1' (i.e. "ignore") will cause LVM mirrors to be ignored and prevent the problem from happening. Ignoring LVM mirrors also means that it is impossible to stack volume groups on LVM mirrors.

The problem is not present with the LVM RAID types, like "raid1". Users are encouraged to use the RAID segment types - especially when attempting to stack volume groups on top of mirrored logical volumes.
Nenad Peric 2013-10-28 14:18:03 UTC Status ON_QA VERIFIED
CC nperic
Tru Huynh 2013-10-29 10:33:53 UTC CC tru
Corey Marthaler 2013-10-31 18:24:17 UTC Blocks 1025456
errata-xmlrpc 2013-11-21 23:28:41 UTC Status VERIFIED CLOSED
Resolution --- ERRATA
Last Closed 2013-11-21 18:28:41 UTC

Back to bug 1009700