EVA to EVA replication over unreliable WAN links

[Adrian Ogden]

Hi
Any experience/info on doing either async or sync between two EVA's over a FC/IP WAN link would be appreciated. Specifically how well/what happens when the link goes down and backup at unexpected times? I heard that a full resync can be triggered if the link misbehaves to much.
Assume the link is between 150 and 300Mbits/sec 100miles low latency.

[Ivan Ferreira]

I have no experience with this, but I think that the major problem will be, besides the resync bandwithd, what if the HOME eva crashes before the resync, and there are too many changes done?

I think that you need a reliable link (or link pair) to make the CA worthless. (just a personal oppinion)

[Vincent Fleming]

When the WAN link goes down the EVA will use the write history log to store any changes you make to the primary site.

If the link is down for an extended period of time, it's possible that you will fill the log - so you need to make the log LUN large enough to withstand any expected outages.

If you run out of log space, it will require a full resync.

We highly suggest that you have multiple WAN links from different vendors to prevent any long outages from affecting your operations.

Regards,

Vince

[David Tapper_1]

Hello!

Log LUN? Is that a new feature in recent VCS/XCS?
I thought the disk group with most free space was used for the write history log?

Regards,
David

[Ivan Ferreira]

We have CA over reliable links. We do not configured a Log LUN (But I think that can be done).

We only configure the DR groups, the VCS does the rest.

[Vincent Fleming]

I'm sorry I typed LUN - that's how I think of it, and that's how it's implemented internally, but as you point out - no it's not a LUN in the classic sense.

You can, however, pick the disk group that it lives in - which is a good idea - make sure it's not in the same disk group as the LUNs that you're replicating.

[Vincent Fleming]

Oh two more things - logs are Vraid1, and picking the group is a new thing - at least I think that's out now... ;-)

[Peter Mattei]

When using WAN links, synch CA is not recommended due to latency and thus performance issues.

Regarding the data replication log file location:
1. Since VCS3.020 and with all XCS versions the log will be placed in a FATA disk group if present, if not in the disk group with the most freespace.
2. It will not be placed in a LUN but in the freespace of the selected disk group.

You would not need to care about the resync the CA LUNs. The EVA will start resynching when the link comes up again and will also assure in order delivery on the remote EVA.

Cheers
Peter