I'm preparing to migrate a CIFS filesystem from one storage pool to another using fs_copy. We want to perform an initial copy and then at cutover refresh the data utilizing checkpoints. What I'm looking for is someone to double-check my process here.
Here is some background info:
CIFS was configured with a virtual data mover
The current CIFS filesystem is named cifs_r5
# 1. Create target file system of exact same size of source & mount ro
nas_fs -name sata_cifs_r5 -type rawfs -create samesize=cifs_r5 pool=clarata_archive storage=APM##########
server_mountpoint server_2 -create /sata_cifs_r5
server_mount server_2 -option ro sata_cifs_r5 /sata_cifs_r5
# 3. To refresh the data once
fs_ckpt cifs_r5 -name src_ckpt2 -Create pool=clarata_archive storage=APM##########
fs_copy -start src_ckpt2 sata_cifs_r5 -fromfs src_ckpt1
# 4. To refresh the data again
fs_ckpt src_ckpt2 -refresh
fs_copy -start src_ckpt2 sata_cifs_r5 -fromfs src_ckpt1
The part I'm not sure about is with the VDM. Isn't the VDM configuration also saved in the filesystem? How do I make sure that the VDM configuration gets migrated and will come back up? Will the CIFS server & shares come back?
The part I'm not sure about is with the VDM. Isn't the VDM configuration also saved in the filesystem?
The information about the VDM is stored in the NASDB The config information for all CIFS servers that are created "on" or "in" a VDM is stored in a small VDM root file system (root_vdm_X)
How do I make sure that the VDM configuration gets migrated and will come back up? Will the CIFS server & shares come back?
Since you arent moving that VDM root file system it will all remain intact Just make sure your path to the data file systems stays the same (like you did when you mount them on the same dir that the old one was)
Once you start that VDM/CIFS it will all look exactly the same as it was
P.S.: The only reason to move that VDM as well would be if you need to completely free up the FC disks to remove them. That would be possible as well with a few more extra steps, where full Replicator is even more helpfull
i wanted to comment on the steps where you perform initial and incremental fs_copy, you need to add "-option convert=no" option otherwise upon completion of the fs_copy process file system will be converted to uxfs but you want it to remain rawfs until your final fs_copy. You are right about VDM root file system containing share/local group information of CIFS server(s), i am eager to find out how that gets migrated.
if he wanted to move the VDM itself ..what is the procedure ?
roughly the same plus fs_copy'ing the VDM rootfs as well
I'd have to try it to work out a exact procedure since you need different commands to take checkpoints of a rootfs Might also require some extra edits
In that case though I would admit that I'm lazy and just setup up proper loopback Replicator sessions for VDM and data file systems through the Celerra Manager GUI and just tell it to failover both
apart from the convert=no here's my quick feedback:
- unless you have a gateway with multiple storage systems attached you dont need to specify storage= - only for the first checkpoint it makes sense telling where the savvol will go - If that VDM only contains this one CIFS server I would instead of disabling the CIFS interface just "shutdown" the VDM by changing it from loaded to mounted - if you have a Replicator license or get your EMC contact to allow you to use one temporarily you could use Replicator proper through the GUI with a lot less effort and downtime
As always, trying the steps out on a small test VDM/CIFS/fs helps to avoid syntax quicks and will make your production move go much faster and smoother. Since you can create multiple VDMs and CIFS servers that only costs you some time and a temp. IP adress / CIFS server name
I assume you arent moving the VDM itself from FC to SATA so if you make sure that your new SATA fs gets mounted under the same path as the old FC one you dont have to worry about the CIFS config
Thanks! That was exactly the information I was looking for.
Dynamox, Thanks for convert=no tip
Rainer, The reason I'm specifying the storage system & pool for the checkpoints is because the FC pool is completely used up by the existing filesystems. So that was the only way I could create checkpoints. I like your suggestion of shutting down the VDM instead of just disabling the CIFS interface; there is only a single CIFS server on the VDM.
The goal of this process is to move the CIFS data to a SATA pool to free up the FC disks to be used as a file system for iSCSI LUNs. I guess after the CIFS file system is copied and I rename the old CIFS filesystem I can start using it for iSCSI LUNs, correct?
In that case though I would admit that I'm lazy and just setup up proper loopback Replicator sessions for VDM and data file systems through the Celerra Manager GUI and just tell it to failover both
I think I will look into enabling Replicator. Getting experience with these commands is very helpful though, too. I will definitely lab this out before doing it in production.
specifying the pool makes sense, but only once since all checkpoints of a certain file system share the same savvol So for the second one or the refreshes there's no need to specify the pool
yes, I think you can just use it for ISCSI then (once it no longer is in a RepV1 relationship)
I guess you arent bothered by the mere 128MB or so that the VDM takes up on the FC disks so it doesnt make sense to move it as well
I will remove the pool specification from my subsequent checkpoint commands. I don't have a problem leaving the VDM on the FC disks.
I've already ran the first fs_copy but didn't specify the convert=no option. So does that mean that my destination fs has already been converted to uxfs? I would look but don't have access to the system at this moment.
Rainer_EMC
6 Operator
•
8645 Posts
625
0
Posted January 21st, 2009 06:00
The information about the VDM is stored in the NASDB
The config information for all CIFS servers that are created "on" or "in" a VDM is stored in a small VDM root file system (root_vdm_X)
Since you arent moving that VDM root file system it will all remain intact
Just make sure your path to the data file systems stays the same
(like you did when you mount them on the same dir that the old one was)
Once you start that VDM/CIFS it will all look exactly the same as it was
P.S.: The only reason to move that VDM as well would be if you need to completely free up the FC disks to remove them.
That would be possible as well with a few more extra steps, where full Replicator is even more helpfull