Administrative failover of data file system only

Question

Hi guys,

I am trying to do an administrative failover of a data file system. So on source Celerra i ran fs_replicate -reverse fs_dest_marketing:cel=NS80 fs_src_marketing

it went through its 10 steps and now fs_src_marketing is read-only and fs_dest_marketing is rw. Since i am not trying to do VDM failover is there anything else i need to do ? I was thinking that it would create a checkpoint of the target file system in order to be able to re-sync back ..but i just ran fs_ckpt fs_dest_marketing -l ...and i don't see any checkpoints. I am not planning to fail-back so are there any other things that need to be cleaned up before i delete the source file system ?

Thanks

nandas · Accepted Answer

Hi dynamox,Sorry for my late response on this. What you are seeing is quite normal and expected. When you are accessing from windows - the owner of the file is as per the SID information on the ACL. Though your UID changes from one Celerra to another (different User mapping database) - but your SID is the same. The ACL is stored on the file and your SID matches with the ACL - thus you can access it and your windows user remains the owner.Hope this is not much confusing.However, if you are using Quota - then it may be an issue since UID/GID information is used for quota checking.Thanks,Sandip

nandas · Answer

Since you want to mainly migrate the data - no VDM replication, no failback etc - I think you are all set now. However, you may create a share on the destination file system or access it through c$ share on the CIFS server and ensure everything is ok - data is accessible and you can write to it. Also check the ACLs on the files & folders on the destination FS as we discussed earlier.

Running fs_replicate from CLI does not create the reference checkpoints - but if you do it from GUI, it would have created the reference checkpoint.

Rainer & Others may provide some more valuable comments and feedback. Till then -
cheers,
Sandip

dynamox · Answer

Thanks Sandip,now this is the weird part that i was afraid was going to happen. i cd into /nas/quota/slot_2/root_vdm_14/fs_dest_marketing ..on NS80 ..and when i look at UID of these test files, they all have UID of 34650. This UID corresponds to my UID in usermapper on CFS14. I just exported usermapper on NS80 and there is no 34650, yet when i look at ownership it correctly displays my AD account ? How can it display this ownership information if UID 34650 is not even in the usermapper database ?

dynamox · Answer

i see, all this time i was assuming file ownership on Celerra is based on UID/GID. I am going to stop pulling my hair with this usermapper stuff

i don't have user quotas, and multi-protocol file system is already on NS80. I am going to press on with my CFS14 migration, starting with some "not so important" CIFS servers

Thank you for your patience and help.

dynamox · Answer

when i tried to do an administrative failover using -reverse option it works ok, but as soon as it fails over to the destination file system ...it immediately starts syncing back from destination to the source file system. I don't want that. Is there a way to simply fail it over ? On the source tried i tried:

fs_replicate -failover -o sync fs_dest_marketing:cel=NS80:if=cge0 fs_src_marketing:if=cge0

and i keep getting:

Error 2211: -o : invalid id specified

I want to perform a graceful, synchronous failover.

Celerra

Administrative failover of data file system only

Was this post helpful?