Red Hat Bugzilla – Bug 201585
vgchange -a n hangs if access to pv's is removed.
Last modified: 2013-02-28 23:04:20 EST
Description of problem:
If all access to the physical devices is lost a vgchange -a n will fail
(actually it will hang). This is supposely expected behavior but it would be
adventagous to have some way to make this complete. I have some scripts that
in the event of a physical device failure start a halt sequence which will
do a vgchange -a n. This halt script will fail (will it will never complete)
because of this hang.
Version-Release number of selected component (if applicable):
Steps to Reproduce:
1. define a vg and activate it.
2. pull the cables to the pvs.
3. run vgchange -a n.
More info can be provided if required however the issue seems to be well
understood. It would be helpful if this just did not hang.
vgchange -a n will fail with io error if PV is removed (but it will not
deactivate VG then).
If it hangs, probably some low level driver is blocking it (access to lvm metadata).
(Tested in RHEL4.5.)
Please if you spot this bug again, reopen this bug and provide debugging output
of vgchange (add -vvvv) and, if possible, task state log (echo