High Availability for File Server based on UCS


#1

Hi All,

We are now running a file service on only one UCS server (also acting as a backup domain controller) in a UCS domain. What can I do to have the HA (High Availability) for our file service? Can someone give us an advice?

Thanks


#2

Hey,

UCS does not provide HA facilities out of the box. Which software and hardware components you chose is also always pretty specific to the HA scenario in question.

If you need a highly available file service then you shoud start by storing the files on a central NAS or SAN device – preferably of the professional variety, e.g. Dell MD series devices or something in that category. Otherwise you’ll end up with the NAS/SAN being the single point of failure. An alternative is to use and set up DRBD on all nodes in the cluster.

Next you can set up multiple UCS servers that mount the storage from the NAS/SAN via mechanisms like NFS or as iSCSI block devices (for the latter you also need HA software on the UCS servers so that the block device is only ever mounted in a single location).

Next install HA software (UCS includes both Pacemaker and Heartbeat in its “unsupported” repositories) on all systems that assigns a single IP address (e.g. x.y.z.253) to the currently active node and that re-assigns that address as soon as the active node is detected dead.

Last create a DNS entry (e.g. “files.yourdomain”) with the aforementioned IP address x.y.z.253.

Now your clients can access the file services via said DNS name “files.yourdomain” and should always end up on the currently active node in the cluster.

Kind regards,
mosu


UCS als ausfallsicherer Fileserver