Bug 133431 - vgremove hang after failed vgremove attempt
Summary: vgremove hang after failed vgremove attempt
Keywords:
Status: CLOSED CURRENTRELEASE
Alias: None
Product: Red Hat Cluster Suite
Classification: Retired
Component: gfs
Version: 4
Hardware: i686
OS: Linux
medium
medium
Target Milestone: ---
Assignee: Christine Caulfield
QA Contact: GFS Bugs
URL:
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2004-09-23 22:49 UTC by Corey Marthaler
Modified: 2010-01-12 02:58 UTC (History)
0 users

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2004-10-29 21:04:11 UTC
Embargoed:


Attachments (Terms of Use)

Description Corey Marthaler 2004-09-23 22:49:14 UTC
Description of problem:
I've seen this quite a few time when bringing up my cluster.

After everyone is in the cman cluster and the clvmd service is in the
run state:

[root@morph-01 root]# cat /proc/cluster/nodes
Node  Votes Exp Sts  Name
   1    1    6   M   morph-01
   2    1    6   M   morph-06
   3    1    6   M   morph-04
   4    1    6   M   morph-03
   5    1    6   M   morph-02
   6    1    6   M   morph-05
[root@morph-01 root]# cat /proc/cluster/services
Service          Name                              GID LID State     Code
Fence Domain:    "default"                           1   2 run       -
[1 3 6 2 5 4]

DLM Lock Space:  "clvmd"                             2   3 run       -
[1 3 2 4 5 6]


I attempt on one of the nodes to vgremove a volume. This causes these
cdrom drive errors along with a failure of the vgremove cmd.

[root@morph-01 root]# vgremove corey
cluster send request failed: Invalid argument

hdc: packet command error: status=0x51 { DriveReady SeekComplete Error }
hdc: packet command error: error=0x54

I then try the remove again and it hangs and the cluster needs to be
rebooted.

I did turn on the cdrom filter in /etc/lvm/lvm.conf and added hdc but
that doesn't seem to help or do anything.
    # Exclude the cdrom drive
    filter = [ "r|/dev/cdrom|hdc" ]

How reproducible:
Sometimes

Comment 1 Corey Marthaler 2004-09-23 23:00:02 UTC
I'm not convinced that the cdrom messages have anything to do with
this bug because even though I always see them right before seeing
this bug I do also see them other times without issue. 

Comment 2 Christine Caulfield 2004-09-24 07:46:51 UTC
yes, the cdrom messages are a red herring. This is a bug introduced by
me fixing a different bug yesterday.

Comment 3 Christine Caulfield 2004-09-24 09:39:02 UTC
it's more complicated than even that. The following checkin fixes
clvmd to cope with more then one VG lock, but there seems to be an LVM
command-line bug in there too. I need to check with agk about that.

Checking in clvmd-cman.c;
/cvs/lvm2/LVM2/daemons/clvmd/clvmd-cman.c,v  <--  clvmd-cman.c
new revision: 1.2; previous revision: 1.1
done
Checking in clvmd-command.c;
/cvs/lvm2/LVM2/daemons/clvmd/clvmd-command.c,v  <--  clvmd-command.c
new revision: 1.3; previous revision: 1.2
done
Checking in clvmd.c;
/cvs/lvm2/LVM2/daemons/clvmd/clvmd.c,v  <--  clvmd.c
new revision: 1.3; previous revision: 1.2
done
Checking in clvmd.h;
/cvs/lvm2/LVM2/daemons/clvmd/clvmd.h,v  <--  clvmd.h
new revision: 1.2; previous revision: 1.1
done
Checking in cnxman-socket.h;
/cvs/lvm2/LVM2/daemons/clvmd/cnxman-socket.h,v  <--  cnxman-socket.h
new revision: 1.3; previous revision: 1.2
done


Comment 4 Christine Caulfield 2004-09-27 07:43:33 UTC
This works for me now. Alasdair has given provisional blessing to the
change, but it's in CVS anyhow.

Comment 5 Alasdair Kergon 2004-09-27 10:38:13 UTC
It looks to be the only line that got missed when the locking lines
were converted to use new definitions, LCK_VG_WRITE.

Comment 6 Alasdair Kergon 2004-09-27 10:38:52 UTC
Will be in LVM2 2.00.25.

Comment 7 Corey Marthaler 2004-10-29 21:04:11 UTC
fix verified.

Comment 8 Kiersten (Kerri) Anderson 2004-11-16 19:08:36 UTC
Updating version to the right level in the defects.  Sorry for the storm.


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