This discussion is locked
-
All replies
-
Helpful answers
-
Jul 22, 2010 3:51 PM in response to richloweby RobertKite,Hey Richard,
It is likely that you are experiencing cache problems.
Check the Activity Monitor on the controller you are using to do your copies. Make sure the amount of inactive memory is no more than 50% of the amount listed as free.
Bob -
Jul 23, 2010 4:10 AM in response to RobertKiteby richlowe,Hi Bob, thanks for the reply. You could be onto something here. The free memory is showing 45MB with inactive memory showing 6.89GB (we have 8GB in total).
Is there an easy way to clear the cache? Restart?
I should also mention that both volumes are hosted by the MDC master, the copy is done on the MDC slave. The MDC slave is showing only 45MB free where as the MDC master is showing 6.29GB free (8GB total).
Message was edited by: RL@ITV -
Jul 23, 2010 4:51 AM in response to richloweby RobertKite,OK Richard,
You need to purge the cache. However, to do this you will need to install xcode as the OS does not have the purge command by default. Xcode will install that tool.
So,
1. Install Xcode
2. Open Terminal and type purge
It may take a bit to complete but you will notice the inactive memory decreasing and performance improving.
You can have Danielle James come up and automate this entire process for you. If you don't have her contact information ping the Xsan boys at Waterloo and they can provide the info.
Cheers,
Bob -
Jul 23, 2010 8:29 AM in response to richloweby Alex Bourn,How are you doing the copy? Are you using the cvcp command? What kind of data? That timing doesn't necessarily seem so bad -
Jul 23, 2010 9:17 AM in response to Alex Bournby richlowe,I have a scheduled unix task set to run within ARD. We are using the following command
cp -R -p /Volumes/XSAN1/* /Volumes/XSAN2
If 3.5TB takes 12 hours to copy, this is a throughput of 81KB/sec. You can get 60MB/sec through gigE.
I would have thought it would be faster using fibre.
I have purged the memory as Bob suggested so I will monitor the next backup and see ho long it takes.
Thanks
Message was edited by: RL@ITV -
Jul 26, 2010 3:41 AM in response to richloweby richlowe,Well after purging the inactive memory the copy still took 13 hours!
Any other ideas? -
Jul 26, 2010 5:41 AM in response to richloweby RobertKite,OK,
The next step would be to initiate multiple copies at the same time.
Check out this article. - http://www.xsanity.com/article.php/20081209162454331
BTW: Just finished a 23 TB copy in 18 hours. -
-
Jul 28, 2010 2:00 AM in response to Alex Bournby richlowe,Thanks for everyone's help on this. I will check out the cvcp command, however I have found that the latest version of Mike Bombich's Carbon Copy Cloner works with xsan volumes and can carry out scheduled incremental backups of volumes by adding a property list file in /Library/LaunchDaemons.
The last incremental backup took 1 hour and 40 minutes.
A result.
Richard -
Aug 10, 2010 9:40 PM in response to richloweby Christopher Collins3,Hey Richard,
Actually your math is wrong. You're getting 80MB/sec, not KB/sec. 3,500,000(MB) / 43200(seconds) = 81 MB/sec. If it were KB, you'd only transfer about 3 and a half GBs in 12 hours.
What kind of storage are you using? Promise or Xserve RAID? Keep in mind a RAID controller on an Xserve RAID only has 80MB/sec of theoretical bandwidth so each LUN would only be able to ever go 80 MB/sec anyway. Also, if you have medium to smaller sized files, that usually produces longer transfers then large files with big blocks of data.
If most of my assumptions are true I really don't see your transfer speed as out of the ordinary. But I am making a lot of assumptions.