VDM Replication and CIFS

No, you dont have to.

Unless you want to be able to access the readonly destination file systems without a failover.
In that case you would create a CIFS server on the physical data mover to access them.
Since in the normal case the destination VDM is readonly you wouldnt be able to create an extra CIFS server there or to start the replicated one

The way it works is that in case you want a failover you failover the VDM and its data file systems.
When you are failing over the VDM its CIFS servers get stopped, the destination VDM goes read/write and the replicated CIFS servers
are started there with exactly the same name, config, shares ...
Only difference being that they will use a different IP address - but they will be telling that to dynamic DNS, WINS and broadcast

The goal here is to provide a DR site and provide users with access to their DFS shares in the event of a disaster.

We are using our Celerra for DFS file shares.

We also do DR testing once a year and we want to provide the DFS shares during that test. This should also test our ability to fail over and use our offsite Celerra.

So, to clarify;

From our source site:

A VDM with 6 CIFS servers and 7 File systems.

I will replicate the VDM (which should include the CIFS servers?)
I will replicate the 7 file systems for that VDM.

During a planned DR test... we will pause replication between source and destination.

Once we have a physical seperation between source and DR site, we want to "promote" the destination Celerra to provide DFS resources for the persons performing their DR testing.

Once testing is done, we need to revert back to a Source -> Destination replication.

SHHWWOO...

Please keep in mind, during a planned DR testing - you should choose "sync" option while failing over from Source to Destination - which is an administrative choice for a planned failover.
Synchronized failover does the following:
i) Takes the source site to a read-only state.
ii) Takes the last of the change data and sends it across the network to the destination site.
iii) Replays those final changes before bringing the destination site to a read/write state.
Synchronized failover results in no data loss, but takes longer to invoke and cannot be performed if the source site is unavailable. It is more suited to a maintenance related failover as part of a failover plan, as both sides must be available to perform synchronized failover.

If you have not already gone through this, I 'll highly recommend to refer the following technical document available on Powerlink or the Documentation CD -

"Replicating Celerra CIFS Environments"

My 2 cents - after a prolonged silence (vacation days)
Regards,
Sandip

Sounds good.
If you are ok with a DR test that will make the source Celera readonly and start the CIFS server on the offsite Celerra than there isnt much you have to do.
Just failover

Of course before you should check if your environment is ok, i.e.
- checking that the offsite Celerra has working DNS, NTP, time ... config so it can reach your domain controllers
- interface names need to be same for the interfaces your CIFS servers are using
- see how your clients do name resolution - if you dont allow dynDNS you probably have to change your DNS manually



yes, IF you created them on the VDM and not on the physical data mover - check with server_cifs


fine



no need - just failover or switchover like the manual explains



fine - of course your "primary" Celerra wont be serving CIFS at that time
It would be difficult to do otherwise, since you cant have two CIFS server running with the same name

Well, we physically seperate our network connections between our DC and our DR site. We bring servers up all the time in our DR site with the same names as our servers in the DC. Unless there is some technical reason with the Celerra, I shouldnt see why we couldnt bring the VDM (Cifs) online R/W at the DR site and keep the VDM (Cifs) online at the Corp. DC.



We will have to build that into our process, but that has been discusses.




Yes, the CIFS were created within the VDM. Or the VDM owns these CIFS, whatever the correct terminlogy would be.



Since we will be doing a controlled stop and start, CAN we simply pause replication or use some other method?





See my notes above about having duplicate names up and running.

Last question:

Can you "failover" a single VDM or is it all or nothing?

For example, I want to create a "bogus" VDM, CIFS and shares, replicate them and then test bringing that VDM up and testing access to those resources.

The bogus VDM and CIFS have their own network interface but share the same network devices. (FSN0 consisting of cge0 and cge1 with cge3 as the failsafe network).

cge2 is dedicated to our another environment.

Thanks guys, you all have been really helpful.

You can failover a single VDM. You do not have to failover all VDMs of one physical datamover.

It will failover using the configured network name. Not the physical device of the source side. So it is depending on the configuration, f.e. it can use a FSN on the source datamover and a single network device on destination datamover

Thanks for your help. I think my questions have all been answered!

Message was edited by:
DHoffman

actually since you have the VDM replicated your CIFS servers are already there
They just arent started since the VDM state is only mounted but not loaded
when you set its state to loaded (requires rw VDM root) they will just start
(see the VDM manual for details)

just try it with an extra test VDM/CIFS - it wont harm your other ones

If you want to peek at your data its perfectly fine to create an extra CIFS server at the physical data mover level on the DR side.
I'm not sure if you can "browse" the VDM owned file systems to create shares in Celerra Manager, but if you use there real
paths like /root_vdm_X/data then you can also create share - or just use the C$ share
Of course these will be readonly as long as the underlying data file system is readonly

This all worked beautifully.

The destination basically came online and the source became R/O and unloaded.

Now, How in the world to I put it back to its original form?

Forgive me on this... I'm not fluent with all the CLI commands so I might need some guidance here.

Note:

The source VDM is Unloaded.
Destination is "OK"

I have set the FS on the destination as R/O and R/W on the source.

Message was edited by:
DHoffman

Thats easy

change your DR VDM state back to mounted

then change the original VDM state from mounted to loaded

(unless you did changes to the CIFS server that you want to sync back - then you first should fs_copy the VDM rootfs back)

Ok, I figured it out.

I'll have to test this a few more times to create our own internal procedures, but overall this was VERY helpful.

Thanks again!!

DHoffman,

Would you be willing to share your DR test procedure/process? I'm in the same boat and it seems like you've run through the gauntlet already.

Chris