2017-10-03 08:31 AM - last edited on 2017-10-12 07:39 AM by GidonMarcus
Hi, We have 2 datacenters and one is primary and second one is for DR. We use snapmirror to replicate the data to the DR and both are running 8.3.2P9. We are planning to upgrade the DR cluster first to 9.2 and after 2-3 weeks upgrade the primary to 9.2.
Here is the question, during the 2-3 weeks where the DR is at higher level and the DR senerio happens, whether we will be able to reverse resync to primary which is a lower level ?
2017-10-12 01:21 AM
hope this helps
2017-10-12 06:27 AM
You will need to upgrade your PORD machine before failback. or to downgrade the DR one.
keep in mind that in some scnerio's you might need to re-baseline, and the downgrade process might not be desirable if new functionality already enabled (as will be outline in the downgrade advisor or release notes)
2017-10-12 07:48 AM
We have a similar situation with a primary/DR data center design where we have mirrors going both directions (i.e. we run some active workloads in our DR site that mirror back to primary) and we like to roll upgrades in our DR site before we roll them in the primary. Anyway, our on-going protection mirrors are all type XDP + async-mirror (version-flexible snapmirrors) which allows for just the scenario you describe. Yaoguang sent along the information you'll need to get these going.
Here's one other link that we found useful when coming up with our design:
Hope that helps.
2017-10-12 08:00 AM - edited 2017-10-12 08:00 AM
i must warn that the version flexible is only flexible on the version side. but have other limitations that need to be validated before going ahead and converting.