Below is copy/paste from OneFS WebAdmin guide. I have gone thorugh multiple failover/failback on my virtual cluster without issue, steps are very simple. One thing that i don't undertand is why after we complete the failback process we do not get rid of the _mirror policy. It did its job, it pushed new/changed data back to primary cluster so why do we need it. Is it an over-site in the documention or there is a good reason to keep this policy around ?
Thank you
**********************************************
Procedure
On the primary cluster, click Data Protection > SyncIQ > Policies .
In the SyncIQ Policies table, in the row for a replication policy, from the Actions column, select Resync-prep.SyncIQ creates a mirror policy for each replication policy on the secondary cluster.SyncIQ names mirror policies according to the following pattern: _mirror
On the secondary cluster, replicate data to the primary cluster by using the mirror policies.You can replicate data either by manually starting the mirror policies or by modifying the mirror policies and specifying a schedule.
Prevent clients from accessing the secondary cluster and then run each mirror policy again.To minimize impact to clients, it is recommended that you wait until client access is low before preventing client access to the cluster.
On the primary cluster, click Data Protection > SyncIQ > Local Targets .
In the SyncIQ Local Targets table, from the Actions column, select Allow Writes for each mirror policy.
On the secondary cluster, click Data Protection > SyncIQ > Policies .
In the SyncIQ Policies table, from the Actions column, select Resync-prep for each mirror policy.
After you finish
Redirect clients to begin accessing the primary cluster.
When the *_mirror policy is created on DR cluster, protection domain is created on your PROD cluster, you can see it if you run
isi_classic domain list
When you're finished and failed back to PROD, the domain is in Writable state. If you delete the *_mirror policy the domain also gets deleted. The next time you failover and run "isi sync recovery resync-prep" to failback, you will have to wait for that domain mark job to run again - it may take long time.
On the other side of the coin, if you decide to keep the mirror policy and the protection domain on your PROD cluster, you will not be able to mv files to the directory that's under the domain, even if it's in "Writable" state
i guess i don't understand what happens in step 8 ? Why do we need to run resync-prep ? My data has been copied from DR back to primary cluster via the mirror policy and now i am just ready to restart my primary --> DR replication. Does step 8 do something to DR cluster location to prepare it to become replication target again (makes it read/only) ?
the mirror policy also leaves a snapshot (at least under 7.0.2.x) that grows like "nether world" (this americans with their censorship of common not harmful words like hell) - in our case several TB in a few weeks and is not needed, since it is the snapshot that gets created on the source of a replication. It is deleted automatically after one year. We started to delete it right after we failed back, to not waste unnecessary clusterspace
i had an SR# with EMC and they said they are useless. Currently we are trying to either use our processtool for DR to automatically delete these snaps or we will create a feature request to let them delete automatically inside OneFS.
the mirror policy also leaves a snapshot (at least under 7.0.2.x) that grows like "nether world" (this americans with their censorship of common not harmful words like ****) - in our case several TB in a few weeks and is not needed, since it is the snapshot that gets created on the source of a replication. It is deleted automatically after one year. We started to delete it right after we failed back, to not waste unnecessary clusterspace
Do you remember the name of the policy, i want to check on my 7.1.x cluster.
I conferred with the subject matter expert for SyncIQ in Support and he stated that the reason we do not recommend removing the mirror policy is it destroys the LIN map being used by it. So, if you were to have to redo the failover process it has to recreate the LIN map for the mirror policy from scratch. This can be a HUGE time penalty when your failover window is tight and there are millions of files to scan in to creating the new LIN map to be used by the new mirror policy once failover is complete.
I hope this answers some of your question, but please let me know if there is any other information that I can get for you! I am more than happy to help.
I am still not following you, maybe you can work through the below steps with me and explain what's happening with these SIQ snapshots and then how the mirror policy is playing its role.
here is a test scenario:
policy name = upgrade_ir
1) I am replicating from primary cluster to secondary cluster. When i check on primary cluster i see snapshot SIQ-3d510f0edc31e18bdb2100dc1441306e-latest and on secondary cluster i see snapshot SIQ-Failover-upgrade_ir-2015-08-11_19-18-35
This is what i consider my normal state.
************
2) i am failing over from primary cluster to secondary cluster. Upon completing that step on primary cluster i still see the same snapshot but on secondary cluster i now see two snapshots:
3) I am failing back from secondary back to primary. I go to primary cluster, select my policy and hit "Resync Prep". At this point i look on primary cluster and see these snapshots:
In reading through your question I have some information that I hope is helpful for you in understanding the mirror policy and snapshots. The mirror policy is left in place to preserve the LIN map for the policy so any future need to failover/failback would faster and more efficient. This will reduce time to recovery in the event of an unforeseen catastrophic failure or planned migration between the clusters.
As for the snapshots, this is what I received from the SyncIQ SME here in Support in reagrds to your question:
All of the SIQ- snaps interact with yet another database used by the active cluster(whether it's Primary/Secondary) to sync up the LIN map over at the side. That is why you need the SIQ- snaps at all times. If you delete these from say the mirror policy referencing them then you have destroyed that mirror policy.
With regards to the snaps he pasted at end of question. [Based on naming convention]
SIQ-0050568f48352c87ca553e2004ae341d-latest --- explained this as part of the SIQ- explanation above.
SIQ-n2isilonpoc-upgrade_ir-2015-08-11_19-51-57 -- This one is a manually enabled option configured by the Primary's policy. It has nothing to do with the mirror. This can be disabled to prevent creating this snap.
>SIQ-n2isilonpoc-upgrade_ir-latest -- This is the same reasoning as above.
If you have any more questions or there is anything I can clarify, please do let me know.
bhalilov1
1 Rookie
•
114 Posts
4650
0
Posted August 10th, 2015 13:00
When the *_mirror policy is created on DR cluster, protection domain is created on your PROD cluster, you can see it if you run
isi_classic domain list
When you're finished and failed back to PROD, the domain is in Writable state. If you delete the *_mirror policy the domain also gets deleted. The next time you failover and run "isi sync recovery resync-prep" to failback, you will have to wait for that domain mark job to run again - it may take long time.
On the other side of the coin, if you decide to keep the mirror policy and the protection domain on your PROD cluster, you will not be able to mv files to the directory that's under the domain, even if it's in "Writable" state