cancel
Showing results for 
Search instead for 
Did you mean: 

Question about backing up vault stores and indexes

Jeffrey_Kusters
Level 3

Hi,

 

I'm planning an implementation of Enterprise Vault at a customer's site. This is the first time I've worked with EV by the way so forgive me if I'm asking for the obvious ;)

 

I'm strugling a bit with the backup strategy. The client runs a HP EVA SAN and they use Mirrorclones to make offline backups of their data (whenever possible). For those of you unfamiliar with mirrorclone technology: a LUN has a mirror LUN which is in sync all the time. When the backup window starts you perform a so called fracture and present the fractured mirrorclone to the backup server. You can then perform an offline backup utilizing the storage area network instead of the LAN. The big plus is that the throughput on the SAN is much higher and you don't have to worry (a lot) about a backup window.

 

Back to Enterprise Vault: in order to perform a consistent backup, I have to stop the EV services. My first plan was to utilize the mirrorclone technology for the LUNs containing the Indexes and the Vault Stores. The problem is that I forgot about the Safety Copy and the archive bit. If I understand correctly EV uses the Archive Bit to determine if archived items have been backed up and the safety copies can be removed. So if I perform a backup from an offline copy of the data, the archive bit is never altered on the original LUN.

 

I've noticed it's possible to turn off the safety copy option completely but I'm figuring they've put this in for a good reason. But just to be sure: it's a bad idea to turn this option off in a production environment, right????

 

I came across this article "http://seer.entsupport.symantec.com/docs/273151.htm" which talks about ignoring the archive bit. Is this something that can help me out???

 

My final, and least attractive option is to drop the mirrorclone approach but this would cause some major problems in terms of finishing the backup within the backup window.

 

What options do I have?????

 

Thanks for your time!

 

Regards,

 

Jeffrey

 

 

1 ACCEPTED SOLUTION

Accepted Solutions

jgrant50
Level 4

You could restart the services in read only mode and then backup up your mirror copy.  Once the backup completes, you can write a post backup script that copies a text file called "IgnoreArchiveBitTrigger.txt" to your active LUN which, once the services are restarted in normal mode (particularly the Storage Service), will reset the archive bit on all of the files in the directory.  See this link for the explanation:

 

http://seer.entsupport.symantec.com/docs/273151.htm

 

I don't use mirror volumes, but I use the methodology described and it works great.

 

Hope this helps.

View solution in original post

5 REPLIES 5

jgrant50
Level 4

You could restart the services in read only mode and then backup up your mirror copy.  Once the backup completes, you can write a post backup script that copies a text file called "IgnoreArchiveBitTrigger.txt" to your active LUN which, once the services are restarted in normal mode (particularly the Storage Service), will reset the archive bit on all of the files in the directory.  See this link for the explanation:

 

http://seer.entsupport.symantec.com/docs/273151.htm

 

I don't use mirror volumes, but I use the methodology described and it works great.

 

Hope this helps.

Liam_Finn1
Level 6
Employee Accredited Certified

I'm sorry for you that you have to work on a HP EVA rather than a real SAN.

 

The EVA's VLUN technology works fine for small environments with just user file storage but the IOPS from the disks are very bad. We moved off an EVA because we could not get the throughput we needed for EV. When doing FSA we flooded the controllers on the EVA and we always ran the system at 75% capacity all because of the disk / controller IOPS needed for our EV environment

 

We had such errors as database backups failing because the controllers on the EVA were maxed

ESE errors on our Journal mail store database again because of Controller and disk IOPS maxing

 

We found that because of the way the HP EVA virtualizes the LUNS (basically everything on the SAN is a METALUN) that all IOPS were shared across all disks rather than being able to dedicate the required IOPS per server we had miserable performance

 

The weekend we moved everything to a real SAN EMC Cx340 series all of our backup and performance issues went away.

On EMC we can select which disks are dedicated to which server so we can calcuate the required IOPS and size correctly. 

 

for our indexes on the EVA we were getting 1500 IOPS on the EMC we are are hitting 6000 IOPS on less spindles. Simular for our SQL databases and even our Exchange databases.

 

Life with a real SAN is so much different.

 

Again sorry you have to work with a wanabe SAN

 

Jeffrey_Kusters
Level 3
I already came across the KB article you're talking about but it's good to know this option is good to use in production environments. I'm going to do some testing. Thanks!!

Jeffrey_Kusters
Level 3
@I don't recall complaining about performance issues with my customer's EVA so I don't really see the rellevance of your reply. But thanks anyway??????
Message Edited by Jeffrey Kusters on 09-12-2008 05:14 AM
Message Edited by Jeffrey Kusters on 09-12-2008 05:15 AM

MichelZ
Level 6
Partner Accredited Certified

Don't know if you've seen it, but there is a nice whitepaper on EV Backups:

 

http://www.symantec.com/content/en/us/enterprise/media/stn/pdfs/Articles/best_practice_for_ev_backup...

 

Definitely worth to check out.

 

Cheers

Michel


cloudficient - EV Migration, creators of EVComplete.