Comprehensive data protection for all workloads
Post Reply
garymansell
Enthusiast
Posts: 26
Liked: never
Joined: Sep 03, 2012 1:31 pm

Crash Consistent Replication - is it OK to use?

Post by garymansell »

I need to replicate a business critical Java Web Application server machine and its separate MS SQL Server host to a DR site.

Unfortunately, when using application aware replication, the initial snapshot of the machines takes 4.5 minutes and during this time the system is almost unusable by the end users - seeing as I would like to replicate this machine hourly, I have hit a major problem.

Switching to Crash Consistent Replication should massively improve this issue and make the plan workable again, as the initial snapshots taken by the replication process are virtually instant - but is this as risky as it sounds?

Presumably a modern OS like Windows Server 2008R2 and 2012R2 should be able to cope with unexpected power outages and MS SQL 2014 must be designed to protect data, so the replica machines should come up fine if they are ever needed, shouldn't they?

I should be able to gain some peace of mind by configuring a SureReplica job to regularly make sure that the replica systems comes up OK

What are people's thoughts on this - am I being completely out of order thinking this is an acceptable solution, or is this commonly how it is done?

Thanks in advance

Gary
PTide
Product Manager
Posts: 6408
Liked: 724 times
Joined: May 19, 2015 1:46 pm
Contact:

Re: Crash Consistent Replication - is it OK to use?

Post by PTide »

Hi,
Presumably a modern OS like Windows Server 2008R2 and 2012R2 should be able to cope with unexpected power outages and MS SQL 2014 must be designed to protect data, so the replica machines should come up fine if they are ever needed, shouldn't they?
They should, however nobody can give you any guarantees on that. First of all I'd investigate why does it take so long to take a snapshot. Another thing to do is to make sure that you really need to replicate the both machines hourly. Lets check if I got it right - the Java server is some sort of a frontend and data does not change a lot during the day. On the other hand the SQL server generates a lot of changes which might be the reason why a snapshot is being taken so long. Is it required to replicate the machines simultaneously? If not then have you considered the following scenario?

- replicate Java server with a separate job once in an hour
- replicate MSSQL server once a day with another replica job when the load is low (at night) but collect transaction logs for replica every hour (or whatever figure fits better).
garymansell
Enthusiast
Posts: 26
Liked: never
Joined: Sep 03, 2012 1:31 pm

Re: Crash Consistent Replication - is it OK to use?

Post by garymansell »

Hi, thanks for getting back to me.

Are you suggesting somehow automatically applying the SQL server transaction logs hourly throughout the day to the (daily) replicated MS SQL Server to keep it in sync with an hourly replicated Application Server? If so, do you have more details as to how I might go about this?

The Java application server has a file vault (10TB), an LDAP system and the Java Application loadpoint that all need to be in sync with the MS SQL database server at any time throughout the day (close would be good enough)

Both machines are 64GB RAM / 6 core - I presume it is the snapping of the 64GB RAM of each VM that causes the initial snapshot to take so long and affect performance of the PROD System whilst it is being taken?

Is there anything I can do to speed up the initial snapshot, or reduce the affect it has on the performance of the PROD System (as the system runs perfectly fine with the snapshot and when deleting the snapshot at the end of the replication)
PTide
Product Manager
Posts: 6408
Liked: 724 times
Joined: May 19, 2015 1:46 pm
Contact:

Re: Crash Consistent Replication - is it OK to use?

Post by PTide »

Both machines are 64GB RAM / 6 core - I presume it is the snapping of the 64GB RAM of each VM
Normally VMware does not take RAM snapshot.
If so, do you have more details as to how I might go about this?
That might be possible with scripting, however think that first you should find out why taking a snapshot takes so long. I guess it's worth contacting support team.

Thanks
Post Reply

Who is online

Users browsing this forum: Google [Bot], Tesla2k and 167 guests