Thorsten and all,
I've tried again to reproduce the issues I experienced on s10 / sc32 -33 core
patch:
--- 8< ---
pub0-node1:/# metaset
Set name = shared-dg, Set number = 1
Host Owner
pub0-node0
pub0-node1 Yes
Driv Dbase
d4 Yes
d5 Yes
ub0-node1:/# date ; time metaset -s shared-dg -df -h pub0-node0 ; date
Wed Oct 14 17:58:39 CEST 2009
real 7m14.837s
user 0m0.294s
sys 0m0.149s
Wed Oct 14 18:05:54 CEST 2009
pub0-node1:/# metaset
Set name = shared-dg, Set number = 1
Host Owner
pub0-node1 Yes
Driv Dbase
d4 Yes
d5 Yes
--- 8< ---
Even though I didn't (knowingly) correct any other config, the problems I ran
into have all vanished - this issue is no longer reproducible on s10/sc32. I
could repeatedly remove a dead node from a metaset, irrespective of
- whether or not it was offline or booted in non-cluster mode
- whether or not [SUCCESS=return] was added to the hosts line of
nsswitch.conf
I can only speculate about what I might have done wrong in the past - could be
that I had somehow managed to mess up the disk-set by interrupting metaset
commands, running metaset -P on the dead node or anything like that.
I will now go back again to snv111/ohac and double check there.
Thorsten, thank you again for your advice, your technical help and your
implicit
moral support! This has been really helpful so far.
Nils
