This post is more than 5 years old
227 Posts
0
2626
Recoverpoint Direct Access Mode and it's writes????
This might be a goofy question, but when in direct access mode, are the writes being written to the replica volume? I ask because when I remove a CG from direct access mode after X amount of time, it doesn't have to do a full resync, so I'm curious how it gets rid of the writes and knows to just replicate the changes if the writes already occurred.
Idan
675 Posts
1
September 12th, 2016 12:00
Hi there,
RP and RP4VMs do write to the replica volumes/VMDKs while in direct IA. We track the changes in the journal of both production and replica volumes as a bitmap (so only metadata is written there). When disabling direct image access, it's a short init and as part of it, only the changes are being replicated.
Hope that helps,
Idan Kentor
RecoverPoint Corporate Systems Engineering
duhaas1
227 Posts
0
September 12th, 2016 12:00
Thanks for the feedback as always. So to be clear:
1) direct access mode enabled
2) changes while testing against the direct access volume with VMware, changes being made to say the VM's on that volume are tracked in the journal volume?
3) testing is complete, direct access mode is disabled, those changes are killed from the journal volume and all changes being tracked on the source side to replicated out to the destination side
forshr
1.1K Posts
1
September 12th, 2016 12:00
They are written instantly and once the access is disabled RP knows which blocks have been changed and subsequently reads the blocks from the Prod side and replicates them.
forshr
1.1K Posts
1
September 12th, 2016 12:00
Nearly!
2) Changes are tracked as metadata in the prod copy journal and the replica copy journal.
3) Writes relating to changed blocks from both sides are replicated.
forshr
1.1K Posts
1
September 12th, 2016 13:00
2) Changes are tracked on both sides as direct image access causes replication to pause at the Prod side.
3) No. Metadata relating to writes at the source and writes changes to the destination are compared and subsequently unique blocks are read and replicated from source to destination. This is why you will see a short init.
duhaas1
227 Posts
0
September 12th, 2016 13:00
2) changes inside what i'm doing with the direct access mode CG you mean correct?
3) also, are you suggesting that changes I make to the Direct Access mode CG is replicated from the destination side back to the source side?
duhaas1
227 Posts
0
September 12th, 2016 13:00
awesome! thanks again forshr it's interesting to hear that changes to the direct access are captured in the journal vs. the replica, they answers my question! thanks again.
duhaas1
227 Posts
0
September 13th, 2016 07:00
Also, why isn't the capacity of the journal on the destination side ever an issue if that's where writes are being kept during Direct Access? Or maybe I'm misunderstanding the different between the "metadata" and the actual "writes"? Sorry to be a pain
forshr
1.1K Posts
0
September 13th, 2016 09:00
Because the writes are written to the actual replica. The metadata relating to these writes is used to update the bitmap in the replica copy journal.