FreeBSD 6.1, VMWare Server , and ESX

[Diceman]

just saw what has happened to bsdforums.org. it's a shame.


Currently running a convulated mess of Freebsd 6.1B4 with a 6.1 Release kernel. i am running it as a virtual machine on vmware server 1.0.5 and it runs great. i am trying to migrate it to an esx server and the hard drive was not found. from what i could find on the net, 6.1 release should contain updated drivers that will let it work on esx, but if not 6.2 release will.

so here i am trying to upgrade to 6.2 release and not having any luck. while make buildworld is running i am getting random seg fault 11's on different modules each time. i decided to just try and move my entire system to 6.1 release since it is kind of a hybrid by accident. im getting the same thing with the 6.1 sources as well.

i did a little bit of research on seg fault 11 during large operations (kernel building, large packages, etc) it appears that it could most likely be a hardware fault, but that makes me wonder since it is virtual hardware. one would think that is the most stable type.

my question is, i have a 6.1 release machine that is on physical hardware and it appears that i was able to download the 6.2 sources and complete a make buildworld. could i nfs mount the /usr/src and /usr/obj diretories to my hybrid machine to make installworld with those sources, or is make buildworld only building modules for the system it was run on? both of the machines have most of the same software but not completely.

to try and troubleshoot why it was seg faulting i deleted /usr/obj and /usr/src and downloaded them fresh but it has done me no good thus far. i ran make buildworld roughly 10 times during my last few hours of work last night and i started a last one right before i left but i have no doubts that it failed.

thoughts?

[vermaden]

Mayber just try other virtual machine for upgrade, VirtualBox/QEMU for example, then go back to your VMware server.

[phoenix]

Yes, you can nfs mount /usr/src and /usr/obj into the VM and run installworld from there. This is the recommended way of doing source upgrades on multiple systems with similar hardware (build on one machine, install on many).

Signal 11 errors are sign of (usually) RAM issues. You should run memtest86 or memtest86+ or similar on the host hardware. It could also be signs of overheating CPU, or dying harddrive.

[Diceman]

Originally Posted by phoenix

Yes, you can nfs mount /usr/src and /usr/obj into the VM and run installworld from there. This is the recommended way of doing source upgrades on multiple systems with similar hardware (build on one machine, install on many).

Signal 11 errors are sign of (usually) RAM issues. You should run memtest86 or memtest86+ or similar on the host hardware. It could also be signs of overheating CPU, or dying harddrive.

awesome, i will give that a shot.

as far as the hardware, i moved this to a temporary piece of gear. it is a brand new in the box 1950. it has error checking ram and has done some significant workload since i started testing/preparing for this migration. i realize that new hardware could still be faulty, it just seems unlikely.

thanks for the answer on the sourcing. hopefully installworld will just complete and i can move on.

if not, im sure you guys will hear from me as i rebuild into a 6.2 or 6.3 VM on the ESX server.

[anomie]

Originally Posted by Diceman

Currently running a convulated mess of Freebsd 6.1B4 with a 6.1 Release kernel. i am running it as a virtual machine on vmware server 1.0.5 and it runs great. i am trying to migrate it to an esx server and the hard drive was not found.

This may sound strange (and the circumstances are certainly different than your own), but FWIW when I installed FBSD 6.3 as a vmware ESX guest, I was getting "drive not found" errors until I installed the FreeBSD Boot Manager.

I know you're migrating and not installing fresh, but the symptoms of your problem are similar enough that it caught my attention.

If, as a last ditch effort before a fresh install (and after you've taken good backups), you'd like to give this a try:

# sysinstall

Configure -> Fdisk -> Q (for quit) -> BootMgr

That should install the boot manager for you, or I believe you can use boot0cfg(8) instead.

Good luck.

[Diceman]

Originally Posted by anomie

This may sound strange (and the circumstances are certainly different than your own), but FWIW when I installed FBSD 6.3 as a vmware ESX guest, I was getting "drive not found" errors until I installed the FreeBSD Boot Manager.

I know you're migrating and not installing fresh, but the symptoms of your problem are similar enough that it caught my attention.

If, as a last ditch effort before a fresh install (and after you've taken good backups), you'd like to give this a try:

# sysinstall

Configure -> Fdisk -> Q (for quit) -> BootMgr

That should install the boot manager for you, or I believe you can use boot0cfg(8) instead.

Good luck.

i know it has a boot manager in it. the problem is that 6.1B4 didnt have the proper scsi drivers to work inside ESX. 6.2 does. supposedly the code for 6.1 to work inside ESX has been "back ported" as far as 4.10 IIRC. thanks for the thought though.

currently working on extracting the 6.2 sources so i can build a kernel.

the nfs mount idea was utter failure, it kept giving me errors about the nfs server not responding fast enough or something like that. i just tar'd up the directories and scp'd them. we'll know soon enough.

so make buildkernel is also failing at random spots. i could probably sit here all night and not get it to work. im wondering if i just need to move this to a different physical server to see if that is causing the issue. problem being that physical servers dont grow on trees. *sigh*

make buildkernel is running through multiple "sections". is there any way to manually go through each section one by one? that way i dont have to test my luck running the entire script each time it fails. it got about half way through the kernel modeles this time before it failed, yet now it has to redo everything before that to try again. im trying to remove the luck factor a little bit so i dont spend the rest of my life trying to migrate this host.


apparently i just had to get pissed and give it a stern talking to, cause make buildkernel just completed. onwards.

i've successfully upgraded the physical machine that i downloaded and ran make buildworld on and it went without a hitch. i have high hopes for the rest of this migration.

thanks for the help guys, and a place to blog my thoughts.