belshaw, john wrote:> I am updating our Lustre config to use Heartbeat for automated failover
> with Lustre 1.4.7.1 and I was checking the operations manual V36 for the
> exact syntax for OST failover and I am not sure if it is correct.
>
> We wish to use active/active failover where 2 OSS see the same set of
> OSS and are active for half of them and passive for the other half.
>
> Up to now we have used failover in a manual way so we have config lines
> :
>
> --add ost --node r909n001 --ost xbb12-1a --failover --lov P-lov --dev
> /dev/sda --size 0 --fstype ext3 --mountfsoptions
"mballoc,extents"
> --mkfsoptions ''-N 1000000''
> --add ost --node r909n002 --ost xbb12-1a --failover --lov P-lov --dev
> /dev/sda --size 0 --fstype ext3 --mountfsoptions
"mballoc,extents"
> --mkfsoptions ''-N 1000000''
>
> For one lun, and we would failover from r909n001 to r909n002 with
> lconf --node r909n002 --select xbb12-1a=r909n002 /etc/lustre/lustre.xml
>
> In the manual I see a different command is used with --service and also
> --group is mentioned in the config file on page 61, however on page 62
> --group is not mentioned so I feel something is missing. I could not
> find a clear answer in the KB either. I would like to use the --group
> option to failover all OST''s with one command. Could someone
clarify the
> syntax for OSS failover?
--service=<ost-name> is the current method. --group and --select are
depreciated. Since you must specify by --service there is no way to fail
all OSTs over as a set. (--group didn''t work that way either, it only
grouped failover pairs.)
cliffw>
>
> John Belshaw
>
> IT technical Architect
> IT Operations - EAME Computing Hub
> CGGVeritas
>
> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss@clusterfs.com
> https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
>