Archive for the ‘hardware’ Category

Scenario:
3-node cluster has access to one disk with GFS/GFS2 filesystem.

  • node1 – offline
  • node2 – offline
  • node3 – online

Node3 waits for other nodes to connect. Unfortunately, node2 and node 3 are phisically damaged.


Problem:
Cluster cannot start and mount GFS/GFS2 filesystem.
Console shows errors:

can't connect to gfs_controld: Connection refused  
can't connect to gfs_controld: Connection refused  
gfs_controld not running  
error mounting lockproto lock_dlm

or

gfs_controld join connect error: Connection refused
error mounting lockproto lock_dlm


Resulution:
Convert GFS/GFS2 cluster to work as a standalone server.
To do this we need to change superblock locking protocol of GFS/GFS2 filesystem.


Check what is current type of locking protocol:

 

gfs_tool sb <device> proto
gfs2_tool sb <device> proto

ex.:
gfs_tool sb /dev/sdb proto
gfs2_tool sb /dev/sdb proto


Change superblock locking protocol to lock_nolock:

gfs_tool sb <device> proto lock_nolock
gfs2_tool sb <device> proto lock_nolock

ex.:
gfs_tool sb /dev/sdb proto lock_nolock
gfs2_tool sb /dev/sdb proto lock_nolock

Restart cluster service or whole server. Disk should mount successfully.
Now you may start to run backup :)


To revert superblock to previous settings, lock_dml, set:

gfs_tool sb <device> proto lock_dlm
gfs2_tool sb <device> proto lock_dlm

ex.:
gfs_tool sb /dev/sdb proto lock_dlm
gfs2_tool sb /dev/sdb proto lock_dlm

 

Helpful commands:

mount -t gfs2 /dev/sdb /mnt/storage
service cman start
service gfs2 start