What is the filesystem that virtual disks reside on the host OS?
Are your guests windows? I'm sure virtual windows machines will slow down over time just like real ones do. I've been using windows VMs for around 6 months, and have not really noticed them slow down, however any new software I want to try gets installed on a new VM, which is then deleted when I've finished using or evaluating the software.
Another thought, have you tried defraging the disks?
1 person found this helpful
I hadn't noticed before that there is no "defragment" option for the virtual disks under server 2 rc 1 as there is under Workstation 6. The shrink option exists under vmware-toolbox, but I don't know how that would affect performance. I know the ws6 manual says that you will get better performance if you preallocate all of the space for the disk when you create it. Another option would be to back up the virtual disk and restore it to a new virtual disk using a backup/restore program that save/restores individual files rather than raw partitions. I know how to do this on linux, but not on windows. You might look to see if the vmware conversion tool could be used to do something like this. I have not looked at it.
The filesystems are Linux/Cetons 5 LVM ext3-fs running on Raid 0 across 5) 750GB SAS 1500RPM Hitachi Drives. I also have 2 WD 80GB Raid0+1 and 1 250gb sas disk.
At one time I was running Windows 2008 Server, (but adhoc.) Most of the main VM-Instances that are running in production are SUSE and RH5. Maybe I should re-install the Ubuntu and RH which seem to give me the most problems.
Has anyone installed VM workstation on the same vmserver and used the utility to vm-fsck or is it even compatiable?
You can't install workstation and server on the same host, though I may have defragmented a disk that I've moved back and forth between server 2 and workstation. Maybe it would work to do this over NFS. I would backup the VM prior to defragmenting. Also, you can't use the shrink on LVM volumes. When I do my installs, I specifiy custom partitioning and just install the boot/root/swap on primary partitions.
I'm not sure what fragmentation issues you may have on ext3. Try noatime, nodiratime mount options.
Convert virtual disks to monolithic flat. They have slightly better performance.
I think it depends on the use of the virtual machine. I have done something like install Fedora 8 or 9 base in a virtual machine, run yum update to apply the large number of updates and then shut the vm down and had workstation 6 popup a message telling me my virtual disk is fragmented. It's possible I may have applied a few more other updates, but the yum update process definitly fragments the file system. It would probably help to put /var/cache/yum on a different filesystem, but sometimes for testing purposes we just need to do a quick install.
I think put yum-cache on another virtual disk does not help much
Bevause this is only the temp-dir for the downloaded packages
Most of the changes while update is running affects the changed files
in the system which will change their size
I have running a fedor9 machine since some days
Host and Guest are Fedora 9 x86_64 and the guest does many updates
of many thousands files in a seperate data-disk with rsync
Until today i see no problems even with the sparse disk (data)
It performs like a charme, since sunday midday i have ruinning an apache benchmarks
In the night all data (4 webservers in the internet) are pulled with rsync
I have there generate times of 0.020 seconds with a php-cms (self-made) under stress
For me it works like a charme, but maybe it will run better in future with ext4-filesystem and
the included support of online-defragmentation
2,702,540,034 querys since sunday
I've just noticed that the vmware-vdiskmanager under server 2 RC1 has defragmentation capability. Personally, I have not been aware of having performance related problems due to disk fragmentation. My only experience has been that I run the defragment when ws6 tells me that it needs to be run. Generally I have avoided using virtual disks for applications such as databases with frequent writes. I'm unsure whether vmware's defragment is defragmenting at the OS filesystem level or at a lower level in their virtual disk layout. fsck under linux will tell you how much fragmentation you have on a filesystem (the non-contiguous parameter).
whitewave 2# fsck -f /dev/sdf2
fsck 1.40.2 (12-Jul-2007)
e2fsck 1.40.2 (12-Jul-2007)
Pass 1: Checking inodes, blocks, and sizes
Pass 2: Checking directory structure
Pass 3: Checking directory connectivity
Pass 4: Checking reference counts
Pass 5: Checking group summary information
u4-sdf2: 44/2518208 files (6.8% non-contiguous), 4560835/5028345 blocks
1 person found this helpful
One thing that comes to mind here is that if you are running vmware's split disk files layered on top of the raid interleaving, you could create a non-optimal disk access pattern. I don't know what kind of access patterns vmware uses for the multifile virtual disks. If you are running raid I think a single contiguous disk file would be likely to perform better (with the best performance being with the preallocated file).