Replcation issue on 7400 (Latency to the Hosts)

[cesarpegado]

we have an issue replicating volumes from 1 3par to another.

we have 2 x 7400 (4 nodes) both set with 2 links sync and 2 link async

we can't use fibre for replication.

Every time i start a sync replicating job up it hammers the connection which causes latency to the host.

How can i get all these jobs replicated and not cause any issue to the hosts?

When i check the ports on the gui it shows the connection running at 100% even though the remote job only has 1 object in it (1.7tb lun) is there a way we can throttle the connection to allow enough bandwidth to do the acknowledgment to the hosts?

So you know if we set all 4 links to sync, will it use all 4 links for the replication? as someone told me that it would still only use 2 links for replication.

Thanks

[Sheldon Smith]

It depends on the version of the HP 3PAR OS you are running on the arrays.

If you are running 3.1.2, you can fill the Primary (which it sounds like you already have),  then make a snapshot and back that up using a block-based raw disk backup program. Take the backup to the Secondary array and restore it to the target Secondary volume.

Finally, back on the Primary array, use the CLI "admitrcopyvv" command to add the Primary VV, its snapshot, and the Secondary VV to the desired Remote Copy Group.

(See the latest HP 3PAR Remote Copy Software User's Guide, chapter 20 (Using Tape for) Initial Synchronization and Backup)

[cesarpegado]

yes we are running 3.1.2 (mu1) but i don't think we have the abaility to backup to tape

[Sheldon Smith]

While the documentation assumes tape, what really matters is being able to do a block-based (not file-based) backup and restore of the snapshot in question. Although tape is typically cheaper in terms of bits/cubic inch, any media that your block-based backup software can use would suffice.

Other than that, as far as I know, the only way to throttle a FC link is to set the speed of the ports on the fiber switches.

[Kenneth Walters]

Curious what sort of IP connection do you have between the two 3PARs?

While syncing run this command: "statrcopy -hb -u m -d 5" what do you see going on?

So the hosts that are part of the synchronous groups have a problem because their writes cannot be acknowledged in time due to network latency. I wonder if you can sync them initially in async then once fully syncd, switch to sync mode?

[cesarpegado]

When you say throttle the connection at the switches. Does that mean when sync does it acklodgements it does this through links to the hosts and not through RCIP ports. Because surely if it uses the RCIP to do the acklodgment then surely you just less bandwith to do it with