[Linux-cluster] Advice on Storage Hardware
Michael Will
mwill at penguincomputing.com
Fri Nov 11 23:28:49 UTC 2005
Interesting. Now what is the point of having a separate supervisor
instead of the two nodes supervising each other? lack of stonith?
Michael
Rick Stevens wrote:
> On Fri, 2005-11-11 at 16:54 -0500, David Brieck Jr. wrote:
>
>> On 11/11/05, Rick Stevens <rstevens at vitalstream.com> wrote:
>>
>>> AC/NC's JetStor dual-SCSI stuff works for us under GFS. We also have
>>> a couple of large SANs (each about 26TB), but the small stuff works on
>>> JetStors.
>>>
>>> - Rick Stevens, Senior Systems Engineer rstevens at vitalstream.com
>>>
>> So you have two host servers connected to the array in a failover
>> configuration? If this isn't the case would you mind explaining how
>> you have it setup to account for a host failing?
>>
>
> Yes, that's how we have it. The client machines mount the primary's
> NFS exports, and the secondary just sits there until needed.
>
> A third "supervisor" machine watches heartbeats and if the primary
> fails, it aliases the secondary's NICs with the IPs of the failed
> primary via an ssh command. When the clients reconnect, they connect
> to the secondary server.
>
> The supervisor can watch multiple 2-server clusters since it's not very
> busy. It ain't pretty, but it works.
>
> ----------------------------------------------------------------------
> - Rick Stevens, Senior Systems Engineer rstevens at vitalstream.com -
> - VitalStream, Inc. http://www.vitalstream.com -
> - -
> - "I'd explain it to you, but your brain might explode." -
> ----------------------------------------------------------------------
>
> --
> Linux-cluster mailing list
> Linux-cluster at redhat.com
> https://www.redhat.com/mailman/listinfo/linux-cluster
>
--
Michael Will
Penguin Computing Corp.
Sales Engineer
415-954-2822
415-954-2899 fx
mwill at penguincomputing.com
More information about the Linux-cluster
mailing list