Unsolved
This post is more than 5 years old
1 Rookie
•
44 Posts
0
1404
October 1st, 2017 03:00
Recoverpoint EX on combination of Gen6 and Gen5 RPAs
Hi Admins,
In our environment, we run Recoverpoint EX 4.4.1 SP1 P1 on a combination of 2 Gen6 and 2 Gen5 RPAs on each site. The backend is VNX 5600 with dedicated mixed pools for production LUNs and SAS pools for Journals. We are running through 13 CGs with a total protected size of 115 TB. Each CG is mapped to a single RPA and there is no distributed group writing through multiple RPAs. We have a 1 Gig WAN link which is 60% utilized most of the time. Finally, the failover / failback operations are orchestrated through VMware Site Recovery Manager 5.5. During our daily health checks, we do not see any LAG more than 50 MB on each CG.
During our recent DR drill of all CGs, I noticed there was a point when an RPA went into high load and returned to normal instantly.
Our RPO and RTO values are higher but with the last DR drill experience (as everything went smoothly) - We are looking at cutting the RPO / RTO values to half. I do not know if SRM plays any role in the RPO / RTO cut down time but please advise if the current setup in our premises needs any improvement?
Question: Rather having 13 CGs, can we combine multiple CGs into one ?
Question: Changing the group policy of each CG to "critical" would make any difference ?
Question: Can we failover all CGs at once ?
Would appreciate your time.
Regards
virtualphoton
1 Rookie
•
44 Posts
0
October 2nd, 2017 00:00
During the failover / failback - I did not see any LAG on WAN link or the journal copies.
forshr
2 Intern
•
1.1K Posts
0
October 2nd, 2017 00:00
Did you notice the lag on the WAN link or the journal copies?
Regards,
Rich
forshr
2 Intern
•
1.1K Posts
0
October 2nd, 2017 04:00
So did the highload occur when you performed the actual failover?
virtualphoton
1 Rookie
•
44 Posts
0
October 2nd, 2017 22:00
Yes - During the actual fail over.
forshr
2 Intern
•
1.1K Posts
0
October 3rd, 2017 02:00
One last question. Did the SRM PG contain all of the RP CGs?
virtualphoton
1 Rookie
•
44 Posts
0
October 3rd, 2017 03:00
No.. Each RP CG contains a set of critical and non-critical applications. Similarly we have segregated the SRM PGs into multiple PGs. I can say it is a 1:1 mapping from RP to SRM.
forshr
2 Intern
•
1.1K Posts
0
October 3rd, 2017 09:00
OK, so going back to your original e-mail, SRM will generate it's own bookmarked snapshot as part of the recovery mode procedure and this snapshot will have to be replicated and then distributed before it can used by SRM. So in respect of the RPO the rate of write throughput versus the size of the WAN link is important to minimize RPO/data lag. Setting the criticality can play a part in providing preferential replication capabilities for those selected CGs. As for RTO, this is dependent on distribution to the journal/replica and the performance associated with it. An indication of slow journal will show as lag in the copy journal.
So if you want minimize both RPO/RTO then sizing is important. Moreover, in relation to the number of CGs and initiating recovery actions using SRM then a balance of CGs across the RPAs is important and the PG to CG relationship. Initiating a PG with multiple CGs will may not necessarily impact the RPA but could impact the journal as the RPA will try to roll the snapshot to the replica for multiple CGs, Where an RPA has to perform this action for multiple CGs then this can cause a highload.
Regards,
Rich
virtualphoton
1 Rookie
•
44 Posts
0
October 3rd, 2017 23:00
So would it be a good idea to run a total protected size of 115 TB (irrespective of whatever the CG size maybe) on a combination of Gen6 and Gen 5 RPAs ?
forshr
2 Intern
•
1.1K Posts
0
October 4th, 2017 05:00
Possibly but using a DCG utilising the maximum of four RPAs. Obviously the effects granularity and you will need to check that that the total write throughput does not exceed what the DCG can handle.