5

I've currently got a storage server running Oracle Solaris Express 11 with 12x1tb disks. It currently provides two functions: 1) Providing a datastore for a small vSphere 4.1 farm over NFS and 2) providing a network share for a team over SMB.

The advantages in ZFS are obvious for the uses we've had so far. However, we're starting to look at where the single points of failure are and investigate eliminating them.

An obvious single point of failure is the NAS itself. If there's a software or hardware issue involving Solaris or the NAS hardware, this takes the entire storage pool offline. We've got the budget to buy another equivalent chassis+disks, but what are my software options for clustering these together and making the storage pool more available?

I've read some stuff around HAStorage+ as a piece of software that can do storage clustering, but can that also cluster the NFS and SMB services across both nodes? Can I expect a solution which implements automatic failover?

Finally, are there any decent guides to how to set this up?

growse
  • 8,020
  • 13
  • 74
  • 115

3 Answers3

2

ZFS isn't a cluster-aware file system unfortunately which limits its use in an instantaneous fail-over server cluster. However if your platform can afford a few seconds outage/pause then perhaps something like NexentaStor can be configured to be fairly highly available for a reasonable cost - it's very simple to setup and manage too.

Chopper3
  • 101,299
  • 9
  • 108
  • 239
2

Unfortunately, I don't think you will be able to cluster them in the normal sense as zfs is not a clustered fs. Look at zfs snapshotting as a way to replicate your data. I believe the commands are zfs send and zfs recv.

gdurham
  • 879
  • 7
  • 10
  • I think this might be the way to go. Looking at resources like http://www.cuddletech.com/blog/pivot/entry.php?id=984 I can replicate a snapshot every minute or so and have good-enough recovery data consistency. I just need a script now that can do the zfs snapshot/send/receive bit over ssh and whack it on a cron. – growse Apr 26 '11 at 14:14
  • Thats how we are doing it to our dr location at this point. Very powerful stuff. – gdurham Apr 26 '11 at 14:30
  • Just playing with it now and having major issues getting the remote ssh user to receive the snapshots. :( Things like 'Cannot receive new filesystem stream: permissioned denied'. Any tips? – growse Apr 26 '11 at 14:40
  • 1
    Do you have a user which has permissions to import a zfs stream? You will need to look at zfs permissions to set this up. Check this website out: http://blog.laspina.ca/. – gdurham Apr 26 '11 at 15:50
0

The filesystem doesn't have to be aware of the cluster. With dual attached disk arrays and some clustering software, the surviving side attaches the failed side disks and then imports the zpool. It also works with UFS or anything else you can stick on a hard disk. I've use the Sun clustering software a few years ago. I believe the oracle "Solaris Cluster" is the same thing.

JOTN
  • 1,747
  • 1
  • 10
  • 12