Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.
Hi Mark: On 9/25/07, Mark Steele <mark at zango.com> wrote: > > I found this on the Lustre wiki: > > http://wiki.lustre.org/index.php?title=Drbd_And_Lustre > > Has anyone tested this out? > Yes, however, your the subject of your e-mail is Lustre in primary/primary mode. I don't mean to assume, however, if you mean the same drbd device shared between two OSS nodes, that would be incorrect. You can however pair OSS machines, as is the only way, and have linux-ha failover the OST in question to the secondary drbd device, making it primary on the stand-by OSS in question. The standby OSS in turn can be primary for it's own OST which fails over to the paired OSS node, so, you end up with an active/active OSS setup. As a small example: OSS-A has 2 drbd devices available, call them resourceA & resourceB. OSS-A is primary for resourceA and secondary for resourceB. OSS-B is primary for resourceB and secondary for resourceA. Pairing failover nodes OSS-A & B would allow you to failover the OST (drbd device) in question from the failed OSS to the stand-by OSS. MDS/MDT's are configured in active/passive mode only as, as of lustre 1.6, we can only have one. There is discussion right now on the lustre mailing list regarding HA which you may be interested in. -Mustafa. -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20070925/f01b8381/attachment.htm>