You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I built kmod rpms following instructions in the web documentation. The build was on a system with kernel 5.14.0-503.16.1.el9_5.x86_64. I saved the rpm's in our repository for later use.
Yesterday I installed a system with kernel 5.14.0-503.16.1.el9_5.x86_64. zfs.ko and spl.ko installed in /lib/modules/5.14.0-503.16.1.el9_5.x86_64/extra/zfs, not the current kernel. Thus the system came up without ZFS
The whole point of kmod is that it is valid for the whole minor release, in this case 9.5. It should install in the current kernel. The current setup makes it difficult to automate installs. I was able to recover manually by copying zfs.ko and spl.ko to the right place and running depmod. But if I install on a system that doesn't have the right old kernel in /lib/modules, I conjecture that I wouldn't get those files anywhere.
Describe how to reproduce the problem
Include any warning/errors/backtraces from the system logs
The text was updated successfully, but these errors were encountered:
either I have to get my eyes checked or you have to re-check this report - as I can't spot any issues
the versions of your post in order (done by not copy each at a time but just deleting everyting else from your post):
5.14.0-503.16.1.el9_5.x86_64
5.14.0-503.16.1.el9_5.x86_64
5.14.0-503.16.1.el9_5.x86_64
5.14.0-503.16.1.el9_5.x86_64
I can't spot any mismatch - please re-check on your end
sorry, I was looking at the system on which the rpm was originally built. It was built on a system running 16.1. The rpm wa installed on a system running 19.1, but it installed into the 16.1 /lib/modules, which was there because Redhat typically keeps several old kernels. So on a reboot ZFS didn't work.
I checked the rpm with rpm2cpio. zfs.ko and spl.ko were in /lib/modules/5.14.0-503.16.1.el9_5.x86_64. Unfortunately I think they need to go into a version-independent place and then get copied or symlinked into either the most recent kernel or all 9.5 kernels. Once it's put in the current or most recent kernel (I'm not sure which), Redhat will symlink it into new kernels as they are downloaded and installed, as long as they are consistent with the ABI of the new kernel.
I'm in the process of adding a script to my ansible setup that reads the rpm directly with rpm2cpio and puts the files into the most recent kernel. But that doesn't seem like something I should have to do.
System information
Describe the problem you're observing
I built kmod rpms following instructions in the web documentation. The build was on a system with kernel 5.14.0-503.16.1.el9_5.x86_64. I saved the rpm's in our repository for later use.
Yesterday I installed a system with kernel 5.14.0-503.16.1.el9_5.x86_64. zfs.ko and spl.ko installed in /lib/modules/5.14.0-503.16.1.el9_5.x86_64/extra/zfs, not the current kernel. Thus the system came up without ZFS
The whole point of kmod is that it is valid for the whole minor release, in this case 9.5. It should install in the current kernel. The current setup makes it difficult to automate installs. I was able to recover manually by copying zfs.ko and spl.ko to the right place and running depmod. But if I install on a system that doesn't have the right old kernel in /lib/modules, I conjecture that I wouldn't get those files anywhere.
Describe how to reproduce the problem
Include any warning/errors/backtraces from the system logs
The text was updated successfully, but these errors were encountered: