app fails to start - XEN interface version check failed?

[jonesy]

I got this error when doing an 'app start' on my grid today:

Building application...
*** configuration of volume z.class.dbase1.boot failed, error log follows:
XEN interface version check failed, this boot volume should be upgraded
Expected ver=30203, found 0
*** configuration of volume z.class.dbase.boot failed, error log follows:
XEN interface version check failed, this boot volume should be upgraded
Expected ver=30203, found 0
Failed to build application 'z' - see log for details.

I did a 'log list' at that time, and got this:

Jan 18 10:53:54 BRZ-VPDC kernel: nbd12: NBD_DISCONNECT
Jan 18 10:53:54 BRZ-VPDC kernel: nbd12: Receive control failed (result -32)
Jan 18 10:53:54 BRZ-VPDC kernel: nbd12: queue cleared
Jan 18 10:53:55 BRZ-VPDC kernel: nbd19: NBD_DISCONNECT
Jan 18 10:53:55 BRZ-VPDC kernel: nbd19: Receive control failed (result -32)
Jan 18 10:53:55 BRZ-VPDC kernel: nbd19: queue cleared
Jan 18 10:53:57 BRZ-VPDC kernel: md: md6 stopped.
Jan 18 10:53:57 BRZ-VPDC kernel: md: unbind<nbd2>
Jan 18 10:53:57 BRZ-VPDC kernel: md: export_rdev(nbd2)
Jan 18 10:53:57 BRZ-VPDC kernel: md: unbind<nbd6>
Jan 18 10:53:57 BRZ-VPDC kernel: md: export_rdev(nbd6)
Jan 18 10:53:58 BRZ-VPDC kernel: nbd6: NBD_DISCONNECT
Jan 18 10:53:58 BRZ-VPDC kernel: nbd6: Receive control failed (result -32)
Jan 18 10:53:58 BRZ-VPDC kernel: nbd6: queue cleared
Jan 18 10:53:59 BRZ-VPDC kernel: nbd2: NBD_DISCONNECT
Jan 18 10:53:59 BRZ-VPDC kernel: nbd2: Receive control failed (result -32)
Jan 18 10:53:59 BRZ-VPDC kernel: nbd2: queue cleared
Jan 18 10:54:00 BRZ-VPDC applogic: [66085:warning s:0]: srv2: AppLogic diagnostic output: reconnection enabled

I did notice a message on the dashboard about vol repair, so I did a 'vol repair --all', but after running that, the same behavior and log entries persisted.

I reported this to support, but it's been 4.5 hours with no response, and I'm unable to start my app, so any clues would be greatly appreciated. Thanks.

[PeterNic]

Brian,

This usually happens if you have older version of the appliance kernel still in the /boot directory of the appliance; please make sure you have the correct kernel and initrd (and properly pointed to from the class General settings -- should match the ones in the LUX5 appliance for example); also make sure that older rev kernels are removed from the /boot directory (there is a bug in the build system that barfs on the old kernel in /boot even if it is not used).

Let me know if this helps,
-- Peter

[PeterNic]

If you haven't just upgraded the appliances and all I said in the previous message sounds like gibberish, then the other likely reason for the failure is that the /boot directory does not contain a valid kernel (possible file system corruption). If this is an instance of a catalog class, 'app clean' and 'app start' should fix the problem. If these are singletons, you will probably need to look at the volumes without the appliance running (sftp to the volume through the grid controller).

Regards,
-- Peter

[jonesy]

Support said something about kernels in /boot, and they fixed it. However, I think I might have an issue with this:

The two singletons that had the problem (out of about 10 components) happen to be the only two that I ran "yum update" on. Are you saying that if I run 'yum update' and there happens to be a kernel update, that's going to invalidate my components? So... I can't upgrade or build custom kernels for my components?

Can you please let me know what the requirements are of kernels and the /boot directory? At some point, there *are* modules I might like to build into my kernels, mostly for performance metrics and such. Is this out of the question? I assumed it was not, since it would appear to be possible to build your own component from scratch using your own OS image! :)

Thanks for your input!
brian.

[PeterNic]

Brian,

You can omit upgrading the kernel when doing the yum update. Appliances use their own kernel and not the standard Linux distro kernel.

You can, however, build kernel modules, apply patches and really do anything you want with the kernel -- simply start from the appliance kernel sources. The download link has been posted a few times on the forum, let me know if you can't find it (make sure it is a link for version 2.1.0 or 2.1.1)

Regards,
-- Peter