VMware Communities
FoxtrotIndia
Enthusiast
Enthusiast

30% CPU load with no VM running

I am running into a strange problem with Fusion 3.0.1 on the latest Snow Leopard.

Sometime Fusion, with all its virtual machines suspended, starts, after a while, running at 50% CPU load. After opening and closing the preferences panel of Fusion, the load dropped to 30%. Any idea what might be causing this?

(process samples for both situation are attached)

Tags (2)
Reply
0 Kudos
7 Replies
rcardona2k
Immortal
Immortal

This is a known bug attributed to an Apple bug in OS X . The issue goes away by quitting and restarting Fusion. In the months I've had Fusion running 24x7, I've run into this 2 times. Not sure what triggers it.

Reply
0 Kudos
raznick
Contributor
Contributor

It was claimed to be an OS X bug, but there were never any real details provided that prove it beyond a single random posting related to a Norton product. I had hoped to hear back about whether anything was submitted to Apple or if there was any further confirmation about the cause of the problem. Unfortunately, I have heard nothing. It's a really frustrating bug.

Reply
0 Kudos
raznick
Contributor
Contributor

Oh, and I'll just add that I run into it every single day.

Reply
0 Kudos
Alfazone
Contributor
Contributor

I have tested it with leopard 10.5.8 and the same happens there as well. It's easy to reproduce the bug, I can just hit check for updates in the fusion menu about 10-15 times and then I can sit and see that cpu and memory usage of the fusion ui grows. This also happens if you don't have any vm and on a reinstalled clean snow leopard and leopard with the only thing installed vmware fusion. It never happened to me with vmware fusion 2.x

Reply
0 Kudos
FoxtrotIndia
Enthusiast
Enthusiast

This is interesting. With my installation as well, when I do a few manual check updates in quick succesion, the CPU load starts to grow, and continues to grow afterwards. I did two test, one with about 20 checks, then the cpu load rose to 100% pretty quickly, and one with 5 checks, then the cpu load rose much slower, after a few minutes it was at 17%, finally rising to over 50%.

Strangely I had the impression that if I did only one update check, the cpu load would not rise, staying constant at 0.1%. A second update check 4 minutes later raised the load to 0.9%, and it keeps rising from then on, after a good 5 minutes it is at 1.5%. I will let it run a bit to see where it stops (100%?).

In all of this I have started Fusion, it loaded the windows of a few suspended VMs, and I did not start any VMs, I just invoked the check for updates command with the mouse. I had automatic update checking disabled. Given how repeatable this problem is, I hope VMWare will soon be able to fix this annoying issue.

Maybe one should disable automatic update checking in the hope that this will prevent VMWare from triggering this bug.

Reply
0 Kudos
rcardona2k
Immortal
Immortal

Check for updates is a good reproduction case! Worked for me getting my CPU to climb from < 1, by 1% until about 15% then big steps past 30%, then level off to 100%. Interestingly, the load switched between cores from sample to sample with nothing else running except Fusion and Activity Monitor.

The implicator in the Activity Monitor samples is the call to -[NSProgressIndicator _animationIdler:]. Which implies avoiding the native Cocoa progress indicator. Fusion 3 uses it's own progress for suspend/resume which may be a way of avoiding NSProgressIndicator.

My usage pattern is that i don't quit Fusion, ever. I do have "check for updates" checked but I never invoke it manually (which brings up the UI and progress bar.) My idle CPU for the Fusion UI is < 1%.

Do you have other use cases which bring up the native OS progress bar or barber pole indicators in Fusion?

Reply
0 Kudos
FoxtrotIndia
Enthusiast
Enthusiast

I have now let this test run for while (Invoked Check for Updates twice on a freshly started Fusion instance, without any VMs running), I have now points at 2.8% utililisation, 9.0% an hour later and 30% another hour later, with it stabilising at around 50% to 60% in the end. Process samples at several points are attached.

What I find interesting is that VMWare seems to consume more and more time sorting (CFQSortArray) as time passes, as if it is filling a list, but not removing items from it once done with them.

Reply
0 Kudos