linux guest: I/O between two drives hangs?

[jackmc]

I am running vmware workstation 6.0.4 build-93057 on windows vista home premium x64 6.0.6001.

I set up a linux guest (Debian 4.0), but I absent-mindedly accepted the default of 8GB for the disk. By the time I realized the mistake, I had finished a lot of configuration on the guest that I did not want to have to repeat with a new install.

I decided instead to add a 32GB disk to the guest and mount it as an extra filesystem. I had no problems until I started to move /home over to the new mount point (via rsync). After copying some of the data, the process just hung. The system had zero load and 99% free CPU, and there was no I/O. Every time I repeated the experiment, I got the same results.

If I instead use rsync to populate the new drive from a filesystem on another machine, then I do not hit this problem. It is only when I try to move data between the two vmware disks that there is a problem.

Both of these disks are "files" on the same physical disk in the host system. While this is not particularly efficient, it should not cause the I/O to hang. It is almost as if the I/O is deadlocking in the host system with both reads and writes to the same physical drive on the host, though they are distinct drives on the guest.

Any thoughts?

[wila]

Hi,

Are you by any chance still logged in as the user that you are moving the home folder from?

Personally I would solve the problem differently.. I'm not sure what type of filesystem you are using but if it is something like ext2/ext3 then it is simple to resize... just take a backup of your VM first.

Then increase the size of your VM disk... boot from a liveCD like for example the ubuntu liveCD (or the sysrec cd), start gparted and resize the whole partition that your system resides on.

I concur that sometimes it is better to add more partitions and there's certainly nothing wrong with moving a home partition to another disk. Done it many times myself. Quite often I will just take a tar.gz backup of the home folder and unzip that at the target while logged in as root (and only root). Nothing beats simplicity

--

Wil

[jackmc]

I am using JFS. While I have had great experience resizing on AIX and even on systems using VxVM, I have never had good experience with resizing linux filesystems. Of course, JFS Is the same JFS as on AIX, so it shouldn't be a problem.

The larger problem is that my disk consists of two partitions: a boot partition and an encrypted partition that contains rootvg. Expanding a VG on an encrypted partition is not easy (I am not even sure if it is possible). What I could do is expand the disk, and then add a second encrypted partition and make it a second PV in rootvg and then rebuild or extend /home.

In the end, however, I am just working around a possible bug in vmware by doing this. It would still leave the fundamental question: why does I/O between two drives in vmware hang?

[wila]

I am using JFS. While I have had great experience resizing on AIX and even on systems using VxVM, I have never had good experience with resizing linux filesystems. Of course, JFS Is the same JFS as on AIX, so it shouldn't be a problem.

Well I suppose you do know what you are doing there... with that background, youŕe not a newbie.

Strange that you are having bad experiences with resizing linux filesystems. I have resized quite a lot of partitions, physical as well as virtual and can't say that I have had real problems there. Well none other then the "my own fault" type of error I mean.

The larger problem is that my disk consists of two partitions: a boot partition and an encrypted partition that contains rootvg. Expanding a VG on an encrypted partition is not easy (I am not even sure if it is possible). What I could do is expand the disk, and then add a second encrypted partition and make it a second PV in rootvg and then rebuild or extend /home.

Is your home partition in such a volume group? That could at least explain some performance downgrade, but it never should stop IO.

In the end, however, I am just working around a possible bug in vmware by doing this. It would still leave the fundamental question: why does I/O between two drives in vmware hang?

I'm afraid I can't answer that question, especially not without any technical info .. if it is reproducable you could take a copy of your VM, then do it again and get some log information out of it. The vmware.log could contain some interesting hints there.

As a return question, why are you still using an old version of workstation? I mean 6.0.4 is old even in the 6.0.x tree and the upgrade to 6.5.1 is free...

--

Wil

[jackmc]

I'll look into the log.

As for the version, I just bought vmware early this year and 6.0.4 was the latest version. The upgrade to 6.5 does not appear to be free; when I tried, they wanted to charge me. Is there some special path to go through that I do not know about?

[wila]

The upgrade from 6.0.x to 6.5 should be free... try this link:

http://www.vmware.com/download/ws/

--

Wil