russellr
Contributor
Contributor

Regular excessive disk activity

Hi,

I'm using Fedora 7 host with Win XP guest on VMware 6. I'm still evaluating VMW but am very likely to buy it when the trial runs out.

Everything about it is impressive.

However, I've just noticed that every 30 seconds (actually it's on the 60th second and the 35th second of each minute) there is between 5 and 20 seconds of disk activity. Typically the activity is 10 seconds long.

The host is quiet (checked with "top") and the guest is quiet (checked with perfmon.msc).

The activity appears to be entirely VMW doing something.

I'm pretty sure this just started in the last few days, but not sure exactly when.

Recent changes are:

  1. I've added shared folders from the host.

  2. I deleted an interim snapshot.

Does anyone know what VMW is doing twice every minute and why it needs to do so much disk activity?

Also, how can I stop it? It pretty much makes the product unusable.

thanks,

RR

0 Kudos
8 Replies
russellr
Contributor
Contributor

Hi,

This problem basically makes VMWare useless.

Following are the details of my investigation. Please Help if you can point me in any useful direction!

Here's the problem: twice every minute VMWare thrashes the disk drive for between 5 and 20 seconds.

It's using about 30% of my disk doing, from what I can determine, absolutely no useful work.

Host is Fedora 7, Guest is Win XP Pro.

To prove that the problem is VMWare, and not the Guest or Host OSes, here's

what I've done:

1. Started "top" on the host.

2. Started "perfmon.msc" on the guest.

3. Waited until both host and guest were idle.

4. Started a clock on the guest to monitor the times.

5. Watched the hard-disk light on my computer.

6. If I shutdown the guest, then the problem doesn't occur.

(This helps prove it's not the host doing something.)

The computer is a brand new notebook with a single disk drive. The single

disk drive is why this thrashing makes VMWare unusable - it takes too much

out of the host and while the problem happens, the guest loses keyboard and

mouse input.

So, here are some times from this morning. I've shown time-of-day when the

event starts and then the number of seconds that the disk was thrashed:

Time

Seconds of disk thrashing

9:54:28

11

9:54:58

9

9:55:28

0

9:55:58

0

9:56:28

12

9:56:58

8

9:57:28

0

9:57:58

0

9:58:29

11

9:58:59

12

These samples show one minute where thrashing occurs alternating with one minute where it doesn't.

Other longer samples I've taken show it's worse than that.

What I need to know is what VMWare thinks it needs to do twice each minute and how I can stop it from doing it.

thanks,

RR

0 Kudos
larstr
Champion
Champion

Why don't you use filemon[/url] in your guest or inotifywatch[/url] in your host to see what process is causing this activity?

Lars

0 Kudos
russellr
Contributor
Contributor

Hi Lars,

Thanks for the suggestions.

Filemon confirms that the guest is doing nothing (well, VMware tools processes do something every 2 seconds, but there's no change when the thrashing occurs).

inotifywatch crashes if I try to watch the whole filesystem, but I've segmented the usage over various areas of the file system and the only IO is to the Virtual Disks.

So, far all the evidence is consistent - VMWare is regularly thrashing the disk for no particular reason.

I'm going to play around with different virtual disk configurations to see if I can pinpoint what might be triggering this.

For example, I currently have a non pre-allocated VD for the guests swap space. I do this because I want to be able to take image backups the only way to skip the junk in the swap is to have it separated out.

Perhaps VMW doesn't like this for some reason. Also, this VM has 512MB RAM allocate. Maybe VMW likes to write this out every 30 seconds?

Any other suggestions are welcome.

regards,

RR

0 Kudos
continuum
Immortal
Immortal

Hmm - I would test with disabled shared-folders next.

Then I would try with

mainmem.useNamedFile = "false"

in the vmx

Do you need support with a recovery problem ? - send a message via skype "sanbarrow"
0 Kudos
larstr
Champion
Champion

RR,

Have you tried using a pre-allocated disk file for the swap area and see if the behaviour is similar? Have you also tried disabling vmware swap, page sharing and memory trimming?

Lars

0 Kudos
russellr
Contributor
Contributor

Hi Lars,

Thanks for the suggestions....the results have been interesting which is why I've taken a while to reply.

Pre-allocating the Swap disk made no difference.

However, when I deleted the Swap disk and told XP to use the C: drive with "system managed" swap file, this seemed to fix the problem.

But, when I started a second VM with XP guest (also with the Swap disk removed and swap on C: drive), the problem came back!

Switching off shared folders made no difference.

So, the current setup which seems to be OK is:

  • I've reserved plenty of RAM for the VM(s)

  • I've told VMW to "Fit all virtuals into reserved RAM"

  • I've left the Swap on C: drive inside the guests

  • Added your suggested setting of mainmem.useNamedFile = "false"

I say "seems to be OK", because I think the problem is still there but much reduced. Currently, I think I'm getting 2-3 seconds of thrashing most minutes.

It's difficult to say for sure because the duration is short.

Anyway, I'll continue to monitor the situation.

I've also just disabled memory trimming, but it doesn't seem to have changed the situation. I'll disable page sharing too, but that would seem to be a very good thing to keep.

Thanks for your assistance, it has really helped! Smiley Happy

regards,

RR

0 Kudos
continuum
Immortal
Immortal

Is the pagefile of your host ok ?

oops - forget that - you are using Linux

Do you need support with a recovery problem ? - send a message via skype "sanbarrow"
0 Kudos
russellr
Contributor
Contributor

Smiley Wink

0 Kudos