what is that you are trying to achieve by using -both_sides option? What is the need here?
in hypermax os you can't have snap and clone concurrently from a given source volume ( it is either snap or a clone); does that ring a bell for your setup?
also there is a -remote option along with symsnapvx if you are trying to do a restore on R1 while R2 is in consistent state. But I can see you are doing establish though..
The R1 side of our SRDF/A pair is our production database while the R2 side is our disaster recovery copy of our production database located thousands of miles away. On the R1 side we periodically take snapshots of the production database in order to create point in time copies from which we can restore from if necessary.
In case of a disaster where we completely lose the production data center where the production database resides, we currently rely on mounting up the R2 side copy of the production data base at the disaster recovery data center in order to continue processing from the disaster recovery data center.
Since the SRDF/A link can be up to 30 seconds behind replicating R1 data to R2 there is some potential data loss possible. In order to avoid possible data loss we currently create a remote clone off of the R2 side as a point in time copy whose contents coincides with one of the R1 snapshots made on the local side. So in case of a disaster, the R2 clone provides us a point in time copy of the database with no data loss due to SRDF/A replication delay.
What we like to do instead is use the SnapVX establish -both_sides option to create both a snapshot off of R1 and a snapshot off of R2 at the same time that will be identical. The idea is that in case of a disaster we would restore the R2 snapshot back to the R2 side in order to continue processing. This would negate the need to use remote clones off of the R2 side all together.
So the real question is can SnapVX establish snapshots on both R1 and R2 at the same time when R1 and R2 are linked using SRDF/A, and if so, what state must the RDF link be in to execute the establish command as our first attempt show below failed?
Thanks for your explanation Thomas. Somehow your last response came to me after I replied above...
There is something I would like to pick up on regarding your current clone process. In the process where you take a clone of the R2 such that it coincides with an R1 snapshot. I guess that means that you clone the DB R1 to another R1 and take a snap of the cloned R2? That way you would have the same PIT on both sides. This would take time though as the data needs to be remotely replicated first. In the mean time the production database has moved on. In a DR, the SRDF/A R2 copy will hold the latest copy of the data. The clone will be somewhat behind that.
It sounds like you want or need a zero data loss extended distance DR capability.
The only way to achieve this would be a cascaded SRDF configuration R1>R21>R2. This is a 3 box solution where the R1>R21 is synch (metro distance) and R21>R2 is SRDF/A.
In the 3 site solution, because the first leg is synchronous, the R21 site will be equal to the R1. If there is an R1 DR, the SRDF/A leg will get the data up to the point of the outage (zero data loss).
Thanks Jason for replying, I can always use another point of view when designing solutions, so yours is much appreciated.
Our current solution is controlled by our application which has the ability to put the database into a READ/ONLY mode
then flushes its data to disk before we activate the R1 VPSnap snapshot, which is then immediately followed by a
RDF Checkpoint of R1>R2 and activation the remote R2 clone. This solution has provided a zero data loss solution that has worked well for us for years.
Since we are now migrating to the VMAX All Flash Array hardware we want to take advantage of the SnapVX technology and use the SnapVX -both_sides option in order to replace the R2 clone we rely on today. Since we are always running in SRDF/A mode the application maybe able to set mode to synchronous while in READ/ONLY mode before establishing the SnapVX -both_sides option and then immediately setting the mode back to asynchronous, but so far I have not gotten that to work. Below is a first attempt using symsnapvx -both_sides in synchronous mode.
Agreed, You have passed your RDF mode issues now. I believe you said in an earlier post that you clone the R1. That may be the issue. There are a few KB articles on the error you have so should be overcome without too much trouble.
I'm just getting back to this after being interrupted by higher priorities, sorry it took so long. I repeated the test, see below, where I put the RDF device group into synchronous mode before executing the symsnapvx establish command. The resulting message is "Cannot use the device for this function because it is a Copy Session source", so then I attempted the symsnapvx establish without the -both_sides option and that worked. So is this message telling me I have a copy session from the R2 side to a clone on the R2 side? If so, how do I identify the copy session so that I can terminate it? If you're still available to setup a call, that would be great!
SKT2
2 Intern
•
1340 Posts
5452
0
Posted May 4th, 2018 08:00
what is that you are trying to achieve by using -both_sides option? What is the need here?
in hypermax os you can't have snap and clone concurrently from a given source volume ( it is either snap or a clone); does that ring a bell for your setup?
also there is a -remote option along with symsnapvx if you are trying to do a restore on R1 while R2 is in consistent state. But I can see you are doing establish though..