Lustre discuss - Apr 2009 - Regarding Lustre file system high availability

If you have 2 OSS''S that target one OST then you have redundancy on
that particular OST related to a failed OSS. But what redundancy do
you have for that OST itself (the SAN,NAS, or simple mount of a local
raid 5/6) fails. I of course do not mean the actual array , since
there would be a redundancy built into that system (raid) . But the
host server for the OST.

On Wed, 2009-04-08 at 21:20 -0400, Christopher Deneen
wrote:
> If you have 2 OSS''S that target one OST then you have redundancy
on
> that particular OST related to a failed OSS.
Yes.
> But what redundancy do
> you have for that OST itself (the SAN,NAS, or simple mount of a local
> raid 5/6) fails.
What kind of failure are you envisioning?  Lustre does expect the block
device it serves as a target to provide it''s own redundancy (if that is
desired -- if you don''t care if you lose an OST, you can make it out of
a single, unraided disk).  Most people use a RAID5/6 device for OSTs.
But you mentioned failure of raid 5/6 in your scenario so you seem to
understand the need for RAID 5/6 on the OST, so it seems you are asking
about something else, which is what I am not understanding.
> I of course do not mean the actual array , since
> there would be a redundancy built into that system (raid) . But the
> host server for the OST.
The host server for the OST is the OSS but you show understanding of OSS
failover in your initial statement, so again, I must be not
understanding what you are asking about.

Can you clarify?

b.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 197 bytes
Desc: This is a digitally signed message part
Url :
http://lists.lustre.org/pipermail/lustre-discuss/attachments/20090409/184962fa/attachment.bin