Monday, April 2, 2007

AIX varyon volume group in multi nodes environment

On environment with multi-nodes running AIX accessing single share fiber channel storage, I am facing a problem, if one node fails, I can not import volume group residing on that physical disk in an other surviving node, although physical volume owned by that node is accessible from surviving node. OS gives out a error can not access volume group descriptor. It make me crazy, because one reason, why I need expensive fiber channel storage is to have flexibility to access volume group of failure node from surviving node. It takes me a while to figure out how it works. When AIX activate volume group, it places a lock on corresponding physical volume, so other node can not activate this volume group. The varyon command always checks the physical volume and refuses to activate that volume group if it is locked by other. It does not matter if the locking node is still living or death. In order to release the lock, we should issue on locking node
varyon -b volume_group_name 
This '-b' flag breaks disk reservations on disks locked as a result of a normal varyonvg command. Because this flag only works on volume group that is already varied on, we should do it on a node that originally owns this volume group to make sure that if that node fails, other can activate the volume group.

No comments: