UNSOLVED
Replay sizes for Vmware LUN's seem excessive
What is everybody's experience with daily replays on a VMware VMFS Lun? I have created a new LUN, formatted VMFS5 and it has one Server 2003 server on it. The server is 8gb is size total. The replays are as follows for the last 5 days. 700mb, 1.6 gb, 1.21gb, 900mb and 900mb. This is using the default daily replay that expires after 1 week.
Don't these replay sizes seem to be excessive? The server is only 8gb total and these replays are 10 - 20% of changed data a day.
I have already done the following to reduce the replays.
1. Moved the windows swap file to a separate disk on a separate lun.
2. Reserved all the system memory in vmware for the guest.
I have other RDM Lun's that are 1tb in size and the daily replays are 500mb - 1gb a day. This seems more normal to me.
I am open to any advice or suggestions on how to make my vmware luns more reasonable. We are trying to setup SRM for our DR site and the replays are so huge we can't keep up with our current pipe.
Odd, but I suppose it could be logging that is creating all the churn.
To move that VM to a 512KB block you'll need to prepare a new storage type. This can be done during production. Create a new volume with this storage type and map it to your ESX systems. When you format it make sure to choose the VMFS with the smallest supported file size, or the whole point may be moot. Once that's done you should have a 512KB block volume in addition to your standard 2MB block volumes. Just do a Storage vMotion over and see what happens!
hallidayr
48 Posts
916
0
Posted November 30th, 2011 11:00
A problem frequently encountered with email servers is giant replays due to small bits of changed data. Since you didn't mention what the WS2003 server does I can only speculate as to why so much is changing.
With mail servers the approach is to use a volume that has a 512KB block size instead of 2 MB. Then, if a 4 KB file is changed, your replay size can shrink as much as 75%.