Start a Conversation

Unsolved

This post is more than 5 years old

3835

December 22nd, 2017 06:00

Journal for SQL in VMWare

Hello,

I need to make some DR tests with a MS SQL server in VMWare (with raw mapping devices).

I configured everything and it works, but since I've to run tests for about 3 days I have a problem of journal filling up.

I saw that if journal fill up I can still use the DR system until log image access fills up and I think this last one is 20% of journal size.

What are the best way to extend the time I can run my tests? Just increase journal size?

On a 3,5TB environment, with a 500GB journal I can survive like 10 hours. Do I have just to increase journal or I can pause transfer?

Thank you

Max

675 Posts

December 31st, 2017 00:00

Hi Max,

That depends; It's correct that 20% of the replica journal is kept for writes while in image access, so if these 20 percent is the issue in terms of write workload on the replica while in IA, the only way to work around it is to increase the journal capacity.

As for the amount of time to test without disrupting replication, that would be the protection window (in your case, 10hours) - when the PW had passed, image access would still be active but replication would be paused. Manually pausing transfer would help with this one but not with the amount of writes done on the replica while in IA.

Another option rather than adding journal would be direct image access - it can be applied to allow to test indefinitely, what it does is to bypass the journal for the actual writes, essentially the flow is:

  a. Enable logged image access

  b. Make sure that's the PiT you want to perform long test with (if not, move forward or backward in time until the the right image is accessed)

  c. Enable Direct IA

Direct IA means that the journal would be wiped. The advantage is that RP still marks the data in the production and replica so that when image access is disabled, it would be a short init - transferring only the differences.

A common use case is to add another copy (remember each copy is independent) and use direct IA on it, this would allow for the first replica copy to stay intact and to serve as the BC/DR/OR use cases without being impacted by this extended image access. It's possible to configure 2 replica copies on a target cluster.

Hope that helps,

Idan Kentor

RecoverPoint Corporate Systems Engineering

@IdanKentor

No Events found!

Top