Data replication on MSA1000 w/volume shadowing... Possible?

[Layne Burleson_1]

Is this possible?
OpenVMS 7.3-2 on Alphaserver DS10

Main objective is to replicate data from one datacenter to another located 2800 cable feet apart without using VMS clustering.

Standalone node at site 1 connected via a short-wave SFP to FC switch with attached MSA1000. FC switch at site 1 connected via a Long-Wave ISL (inter-site link dedicated fiber) to another FC switch at site 2 which has also has an attached MSA1000.

Can the standalone node at site 1 see the drives at site 2 and use volume shadowing to replicate the data? Again, this is a standalone system at site one with no current system at site 2.

Any ideas or opinions would be helpful.

[Bill Hall]

Sounds like the SAN is configured to allow you to do what you want. It does sound reasonable. Just get the volume shadowing license for the DS10, create and present the LUNS on the second MSA, create and mount the shadow sets. Done ;-).

Bill

[Verne Britton]

I'll let others with SAN experience comment on your proposal (at a glance it sounds OK to me) ... my interest is in your larger plan ... with no VMS system at the 2nd site, what can you (will you) do with that data/disks :-)


Verne

[Layne Burleson_1]

well... actually I have a server at the 2nd site so I do have the capacity to use the data should I ever need to use it. My main goal is to find out if this configuration will work.

[Hoff · Accepted solution]

Clustering and volume shadowing includes the fixes and features that you will encounter and will want here. Clustering and volume shadowing is certainly the most expensive. But it does work, and it works quite well.

Drive level (block-level) replication can also work -- having had repeated bad experiences with this approach over the last twenty-some years, drive-level replication hasn't tended to be my favored solution.

In this case, the storage hardware guys do have good gear, and a write-only controller-level replicated remote storage case is the easiest case. Where these non-cluster cluster schemes tend to tip over with ugliness is usually with shared write access. That's a very hard problem for hardware to deal with.

Host-based volume shadowing (HBVS) also works here.

Getting the replicated server going (and a network link for command, control and cluster communications) would be high on my priority list here. Having that server makes for far easier fail-over, can actively assist with many of the processing tasks, and the secondary server -- if no other task is selected -- can also verify the integrity of the replicated data, and can perform off-line archival processing. (Replicated data isn't a viable defense against application-level and volume-structure-level errors, nor against PBKAC-class ("user") errors.)

But yes, what you have should work -- you have to FC storage controllers here, HBVS can shadow across two FC SAN controllers, and I'd not expect a kilometer of fibre cabling to have any noticeable affect on I/O performance.

[labadie_1]

If you already have Rdb installed, it costs nothing to use Hot Standby

see this mirror of the foc
http://pi-net.dyndns.org/docs/rdb/hotstandby/guide/title.htm

[Layne Burleson_1]

After consulting with someone in HP's services group, it has been determined that this will work. Thanks to everyone's post.