This post is more than 5 years old
40 Posts
0
1532
What is the proper Celerra Replicator Failover?
High Overview: I have a file system for user Home Directories. Apparently, there has been some "slowness" during Logoffs. Roaming profiles are used with our users. I just migrated from Physical Servers to the Celerra. I didn't think it was Celerra related until I opened Celerra Monitor and noticed that starting at 4:00 p.m. - 6:00 p.m. The File System is "Hit" with about 1200 - 2000 IOPS. Which makes sense since the user community (800 Users) are logging off and Roaming Profiles are synchronizing.
I noticed the File System was created from 3 volumes within 2 Raid Groups (Raid 5 4+1). The Raid Groups are composed of 600 GB 10K Luns. Hence, both Raid Groups combined will offer about 1200 IOPS.
I want to perform the following:
I will create a Storage Pool composed of (3) Raid 5 (4+1) Groups. The hard drives are going to be 300GB 15K FC Drives (180*15 = 2700 IOPS).
The Current File System is mounted under a VDM. I will replicate to a new File System composed from the NEW Storage Pool (Which will be dedicated to HOME directories ONLY). Mount it under the same VDM.
Once Synchronized:
#1 Is it as easy as hitting the failover Button?
#2 Will I have to recreate all user shares?
#3 Will it automatically use the interface assigned to the CIFS server that is mounted under the VDM?
I will test but just wanted to get some general feedback.
Thanks Everyone,
Damian
gbarretoxx1
366 Posts
0
October 22nd, 2010 10:00
Hi,
#1 Is it as easy as hitting the failover Button?
Yes. Simply create a loopback replication between your original fs and the new one, then switchover the replication session.
#2 Will I have to recreate all user shares?
No. You can unmount the original fs from it's mountpoint, then mount the new one on the original mountpoint.
Example :
# server_umount VDM -p fs01
# server_mount VDM fs01_new /fs01
You will notice a 1-2 minutes downtime, or even less.
Since the shares are based on the mountpoint, and not the file system name, you don't need to recreate your shares.
#3 Will it automatically use the interface assigned to the CIFS server that is mounted under the VDM?
Yes. You don't need to worry about it.
Obs.: if you want the best file system layout in your case would be to use 4 X R5 4+1, then stripe your FS across these 20 spindles. This is what AVM tries to do when there are available disks. This also will give you a future grow room.
Gustavo Barreto.
Rainer_EMC
8.6K Posts
1
October 23rd, 2010 04:00
as always I would suggest to create a test VDM / CIFS and two test fs to try this and figure out the best order of commands before doing it on production data
idsdtapia
40 Posts
0
October 23rd, 2010 17:00
Original config; I did create (4) Raid 5 4+1 Raid Groups out of 600 GB 10K Drives. Hence, I was using 20 drives. However, when I look in Celerra Monitor under Configuration Status, I see two slices existing on two different Raid Groups???
idsdtapia
40 Posts
0
October 23rd, 2010 17:00
Will the failover take care of Step #2? Meaning, If I failover. I will not have to dismount and mount or will I?
Thanks Gustavo,
Damian
idsdtapia
40 Posts
0
October 23rd, 2010 17:00
I do plan on testing early next week.
Thanks Rainer,
Damian
Rainer_EMC
8.6K Posts
1
October 23rd, 2010 18:00
no - failing over the fs will NOT change the paths for the shares in the CIFS config
you could create new shares with sharedup or manually but remounting the fs is the easier way
Rainer_EMC
8.6K Posts
1
October 23rd, 2010 18:00
drill down using Celerra Manager or /nas/tools/.whereisfs to find out where your fs really is
looking at the Celerra MVM and AVM manuals will give you a better understanding
idsdtapia
40 Posts
0
October 25th, 2010 12:00
So I did the following as Test #1:
#1 Replicated a VDM
#2 Replicated the File System and mounted it under the replicated VDM
**** Apparently I can't mount a replica (ongoing) under the the same VDM ****
#3 I unmounted all shares
#4 I failed over the VDM
#5 I failed over the File System
The Replica did become R/W and the replicated VDM did Mount.
I noticed the CIFS server disappeared.
Is this normal?
idsdtapia
40 Posts
0
October 26th, 2010 13:00
Rainer did the following:
#1. I copy my user data to a different file system and created (2) shares.
#2. I replicated the file system.
#3 I paused replication on the VDM
#4 I failed over the VDM
*** At this point I notice, the original VDM unmounts and the replica mounts ***
*** I also notice the CIFS server disappears from Celerra Manager ***
*** It doesn't disappear from AD ***
#5 I attempt to add the CIFS server to AD but it fails that it already exists.
*** I then notice the CIFS server is back in Celerra Manager ***
#6 I pause replication on the file system
#7 I fail over the file system
*** At this point I notice, the Replica (FS) is R/W and mounted under the Replica VDM I just previously failed over ***
*** The original File System is R/O ***
#8 I manage the CIFS server and all shares are still intact with proper permissions
Looks Good.
Any Feedback would greatly be appreciated.
Thanks,
Damian
Rainer_EMC
8.6K Posts
0
October 26th, 2010 15:00
Hi Damian,
sounds good. A bit more complicated than necessary but as long as it got you what you want.
Rainer