VMWare server and linux kernel 2.6.23 - upgrade to latest kernel as fix?

Forums › Linux › VMWare server and linux kernel 2.6.23 - upgrade to latest kernel as fix?

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#23455 30-Jun-2008 09:34

Hi all,

Anyone here had the issues with VMWare server and linux kernel 2.6.23? (This is running on a debian machine)

Even the installer seems to be doing odd things (deleting a required file so the the init.d script complains that an integer was expected).

Ignoring the compilation issues, I'm planning on simply upgrading to the latest linux kernel - so my question is has anybody else experienced good things (tm) by going to 2.6.25.9?

Or am I going to have to revert back to the latest kernel that worked (2.6.18 on my system)

Cheers,
Warren.

logicalit

74 posts

Master Geek

#141534 30-Jun-2008 15:29

Hi-ho,

I'm using vmware server 1.0.6 on , 2.6.18-6 (Stock debian etch) 2.6.24-9 (Stock ubuntu) and 2.6.25-7(custom ubuntu) without any vmware problems.
(lots of weird video problems for the 2.6.25-7 one though, can't get the .config right...

)

Is the vmware server a standalone virtual server, or a workstation as well? There's no good reason to run a non-stock kernel on a dedicated vmware server that I can think of, unless you've got some obscure hardware support issues.

For that matter, there are not too many good reasons for running non-stock kernels on most current distros, unless you like the excitement of it all, or have the afford mentioned wieird hardware support issues...

Cheers, Chris H.

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#141538 30-Jun-2008 15:37

Hardware support - that's why I'm going that way (the MB sensors are not working at all under 2.6.18 ...)

I'm going to try a revert back to 2.6.18 and re-install and check that the installer works AOK.

I'm having an odd problem where the installation routine appears to be deleteing "vmmonInUse" (or similar) - not sure if that's due to the kernel issues reported elsewhere or whether I've well and truly botched my install.

This particular machine is both a server and a workstation ... although "workstation" in the sense that it advertises gdm locally and on the network so that I have something to look at webpages through if I don't feel like using something else.

If the installer works OK under 2.6.18, I might try 2.6.25.

Cheers.

logicalit

74 posts

Master Geek

#141544 30-Jun-2008 15:53

Weird, the vmware installer has always been good to me... (Touch wood)

How did you build your kernel and headers? make-kpkg?

The only odd things I've had happen is with manually built kernels where the headers confused the compile of the vmware modules...

Also, I've seem some how-to's on the web suggesting that you need the vmware patch "vmware-any-any-update-116" or something-or-other for all installs, which is not correct with 1.0.6 from my experience, it works just fine by itself, appart from an issue with shared libraries on ubuntu for the vmware admin tools..

Cheers, Chris H.

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#141639 30-Jun-2008 19:32

Seems there's definitely something odd about the kernel release - I've installed the beta no problems (under 2.6.18), but I think I'll go back to release 1.0.6.

Next step, try the latest linux kernel.

Thanks.

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#141690 30-Jun-2008 21:29

Linux 2.6.25.9 has done the trick - I obviously simply chose the wrong kernel to go to (for those that are curious, I was having issues with getting lvm + a custom kernel going, hence the older version)

So now I have vmware, lvm and my sensors going.

Of course, the other issue I had (the warning about "rtc: lost some interrupts at 1024Hz.") seems to have gone, although replaced with another warning from the rtc about lost interupts.

This *may* be due more to some changes in the kernel however (I shall investigate a little further down the track)

I haven't had to install the "any-any" patch (which I was trying to avoid, as during the next upgrade round how do I what will fail)

I also went back to 1.0.6, rather than 2.0beta, as the beta version of the console appears to want to run up firefox :-/

Regards!

logicalit

74 posts

Master Geek

#141732 30-Jun-2008 23:16

You getting the lost timer interrupts on the client OS's or the server?

The lost timer interrupts on client VM's is a known issue, I've got some notes on it in my wiki at home if that's the problem you're having I'll dig it out...

What video card have you got out of curiosity? I've got an Intel i810 chipset M/B that I can't get the video going with 2.6.25.. A functioning .config file would be nice. :-).

I've not tried Beta 2.0 yet, I'll wait for them to roll it out me thinks, as I rely on my VM's to be up too much to mess about on the fringe, as it were.

Cheers, Chris H.

Dell

Shop now for Dell laptops and other devices (affiliate link).

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#141758 1-Jul-2008 07:35

The server is claiming lost interupts. It's not quite as bad as before, since switching to the graphical login, I'm no longer on the console, so the messages don't interupt my flow.

For the video though, I'm simply using fbdev. I'm also using Xorg rather than Xfree, but I'll put the config file for that up as well. Next on the list is to get postifx working correctly ....

Regards,
Warren.

logicalit

74 posts

Master Geek

#141796 1-Jul-2008 09:38

Try #defoptions= clock=pit in your grub menu.lst, fixed lost interrupts on one box for me a while ago....

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#142023 1-Jul-2008 20:28

Hrm - that seems to be an option more for linux guests, which doesn't really help me much, as I imagine that linux will at most make up about 1/3 of the guest installs I'd be doing. (Doing a lot of installs to check different options, and I'm tired of the testing of different things and the testing breaking my base machine)

Will update this once I sort out the issues.

logicalit

74 posts

Master Geek

#142081 2-Jul-2008 00:07

Ya, it is an option you normally use for linux clients...

I had a problem with a older PIII M/B based host where I was getting 1-2 hours a day slippage, and that fixed it. I was grasping at straws after wasting 5 or 6 hours on a job I quoted 2 for, you're probably familiar with that concept...

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#142140 2-Jul-2008 09:50

What about running NTP? Would that have helped the issue out?

I tend to run NTP on the hosts, syncing with 3 -4 other machines, and then only run NTP clients on the guests.

Lenovo

Shop now for Lenovo laptops and other devices (affiliate link).

logicalit

74 posts

Master Geek

#142153 2-Jul-2008 10:28

I use NTP on all machines, but it may not fix time slew problems with client OS's, for the host it's normally fine.

ntpd only works for small offsets, and you really need the kernel software clock to be vaguely accurate. In the worst case a linux vm client OS can run at half speed, which ntpd cannot correct. (It's also really weird to diagnose the first time you see it!!)

There are a couple of known issues with linux client OS's on VMware, which you normally have to fix using the vmware client package, timesync setting in the VMX file and the clock=pit setting.. It's in the time sync PDF file on the vmware site.

Cheers, Chris H.

wazzageek

1095 posts

Uber Geek
+1 received by user: 108

ID Verified

Trusted

Lifetime subscriber

#149132 18-Jul-2008 12:38

UPDATE; I'm now running kernel 2.6.25.9 ... a lot of my issues seem to have been minimised.

Things I discovered:

- Use the 1000HZ timing
- Don't allow the processor speed to be changed (doh!)
- mostly APM seems ok.

Other than that - in terms of getting the time right inside the guest OS, it seems I need to do an ntpdate first, then run up ntp as the guests seem to start ~ 17 seconds faster than "real time". ntp after that kicks in and things seem to remain stable.

Cheers all,
Warren.