VMware Cloud Community
farri304
Contributor
Contributor

vMotion Fails on W2K3 32bit Guests

Ok this is a strange one that has me baffled.

I have an ESXi 4.1 u1 cluster of 14 hosts with approximately 200+ guests which include W2K8, SLES, RHEL, 32bit, 64bit, and Netware. All of our guests have no issue vMotioning except our Windows 2003 32bit guests.

When I migrate, or throw a host into maintenance mode, the W2K3 guests sit at 10% until (here is the strange part) I open a console to the vm and generate any kind of activity on the guest. This usually means just wiggling the mouse a bit and boom it starts the vMotion.

After that the guest has no problem vmotioning to another host. I notice though if I let it sit for a long time with little or no activity its almost as if the OS goes into sleep mode. At that point I cannot vMotion the guest over to another host without opening a console and generating some activity on the VM.

Since it DOES vMotion I think I can rule out host networking. I think it has something to do with the OS going into sleep mode and possible not being able to communicate witht the vmtools service.

Note: all tools are up to date, all vm's at HWv7, all hosts and vcenter patched to latest level.

Thanks

Twitter: @joefarri
0 Kudos
3 Replies
dquintana
Virtuoso
Virtuoso

Hello Farri,

Just to confirm, do you have a correct configuration un you vm about the nic card (like vmxnet ) driver? second, do you have some irregular in you system and app log into the vm?

Don't forget award points / no olvides puntar las respuestas.

---
Diego Quintana

LinkedIn Twitter WordPress
Ing. Diego Quintana - VMware Communities Moderator - Co Founder & CEO at Wetcom Group - vEXPERT From 2010 to 2020- VCP, VSP, VTSP, VAC - Twitter: @daquintana - Blog: http://www.wetcom.com-blog & http://www.diegoquintana.net - Enjoy the vmware communities !!!

0 Kudos
farri304
Contributor
Contributor

Most of my W2K3 32bit guests are using the Flexible network adaptors. Since vmtools is installed and running on all these guests I thought it automatically used the vmxnet adaptor. I did however try removing 3 of the guests flexible adaptors and replacing them with the vmxnet 2 adaptors and I am still getting the same issues.

Twitter: @joefarri
0 Kudos
farri304
Contributor
Contributor

So I finally figured this out. It seems as though my W2K3 32bit guests were generating too many logs at too small a size. Which was screwing up the vmotion process.

I went into the datastore to view the guest logs and noticed, in some cases, up to 50 different vmware#.log files. I thought this may be an issue so I disabled logging on the guest(edit Settings, Options, General, uncheck Enable Logging), tried the vmotion again and it worked.

I wanted to take a look at how it was generating the logs so I shut down one of my guests and took a look at the config file and noticed the "log.rotateSize" was set to 500bytes. I also had no limit on the "log.keepOld" settings so it was keeping every single log file that with a max size of 500bytes.

Each time a vmotion occurs a new vmware.log file gets created and, I'm assuming, because of the small size and lack of limit on how many it retained it was generating too many new logs at too small a size. Once I changed these settings, based on KB 8182749, everything was working smoothly again. The only issue of course is editing all of those guest config files and scheduling the downtime.

This has been a long time problem for me and I am pretty pumped that its finally resolved. Hope its helpful to anyone else.

Joe

Twitter: @joefarri
0 Kudos