Operating System - OpenVMS
1748219 Members
4619 Online
108759 Solutions
New Discussion

"Enhanced" Asynchronous Replication between EVA's

 
The Brit
Honored Contributor

"Enhanced" Asynchronous Replication between EVA's

Hi Guys,

This is really a storage/SAN/DR question, however I am sneaking it in here based on the fact that I am replicating OpenVMS Data.

Basically, does anyone out there have any familiarity with this topic. We are researching the idea of using this product to replicate data asynchronously between the East Coast, and Reno Nevada. We are concerned about what happens when (or if) the inter-site link fails.

The general understanding is that at this point, the product starts logging (read 'caching') the IO (in a file) until the link is restored, at which time the software goes into a "fast-copy" (done in Synchronous Mode,) until the data is synched up again, and then reverts back to asynchronous replication again.

The problem is "What happens if the Log File is full, and the link is still down??"

Clearly, if your replicated data is for backup/Disaster Recovery, then you would want the software to abandon the attempts to replicate in favour of keeping your production systems running without degradation. And here is the crux of our problem, The product documentation hints or implies that the product will be well behaved, without actually, explicitly confirming the sequence of events.

Previous experiences (albeit with older VCS code) showed that when the link was lost, and the log filled, then local host IO to the local storage sub-system locked up.
Clearly this can be avoided if the log file is made sufficiently large (in the extreme case = the size of the DR group). Our question is simply - Is this the case??

We are not necessarily going to make our decision based on the answer, we are just trying to be aware of all of the implications before we configure and impliment this solution.

We would love to hear from anyone who has implemented "Enhanced Asynchronous Replication" and has any experiences with either testing its performance during simulated link outages, or real outages. Also anyone who is utilizing this with large log files (approaching DR set size)

thanks

Dave