Install zfs-fuse and libguestfs-zfs ( which is required but isn't by default ). Verify zfs is enabled in systemd and running Use zpool to setup a pool mount the pool reboot, everything works. Install a new kernel zfs no longer runs. Verify zfs is running use zpool -import -f <poolname> Now everything works Use dracut -f reboot Again zpool must be run to import the pool. I don't expect there are many people using this, so I understand if it takes awhile to fix.
You have this as a Cauldron bug. Can you please say whether it does not happen with Mageia 7? You say "Install a new kernel", but since 'dracut -f' produces the problem, that points rather to initrd than the kernel per se. Do you agree?
CC: (none) => lewyssmith
(In reply to james Whitby from comment #0) > Install zfs-fuse and libguestfs-zfs (which is required but isn't by default). I should have asked this above. Looking just at Mageia 7, I see no 'requires' by zfs-fuse for libguestfs-zfs; nor that libguestfs-zfs is required by anything. So were you aware of this missing dependency on Mageia 7 also?
(In reply to Lewis Smith from comment #2) > (In reply to james Whitby from comment #0) > > Install zfs-fuse and libguestfs-zfs (which is required but isn't by default). > I should have asked this above. Looking just at Mageia 7, I see no > 'requires' by zfs-fuse for libguestfs-zfs; nor that libguestfs-zfs is > required by anything. > So were you aware of this missing dependency on Mageia 7 also? No.
(In reply to Lewis Smith from comment #1) > You have this as a Cauldron bug. Can you please say whether it does not > happen with Mageia 7? > You say "Install a new kernel", but since 'dracut -f' produces the problem, > that points rather to initrd than the kernel per se. Do you agree? No idea. I suspect you are correct, but I don't know.
Rather try to reply to comments, I'll explain here. I have not tried zfs on m7. I don't have enough space to load another vm to test with. I can tell you for sure that both programs are required for zfs to function. One supplies the base operations , the other the utils to make it usable ( mounting, unmounting etc ).
Created attachment 11667 [details] zfs info before and after a reboot Things have gotten worse. Zfs will not even survive a reboot now. Many updates, no jernel or glibc or anything requiring a reboot. I have attached a file that shows what, when, etc.
Thank you for the attached information, and sorry to have been slow to respond. Assigning to Thierry as the active libguestfs maintainer, and with DavidG (CC'd) for zfs-fuse.
CC: lewyssmith => geiger.david68210Assignee: bugsquad => thierry.vignaud
Don't worry about being slow, I'm sure zfs isn't used by lots of users, I'm just trying it out. I would like to have it work "outta the box",but, again, I'm not in a rush.
FWIW: I had to do a complete install of cauldron: could not load stage t2: now way to save debug info. Anyway. Zfs is functioning properly at boot. It has passed an inrd creation ( instlled vbox guest modules ). Do not know about new kernel... yet.
Suspect there is a problem during shutdown. If I only login/out as root. reboot, etc. ZFS seems to be fine. The "mirrored" disks are used for mounting /home. If I login/out as a user the pool is not remouned on restart. E.G.: Boot, login as root, everything is ok. logout, login as a suer , its still ok. As a user reboot, failure on boot. The pool must be imported again. I suspect because there isn't time for the drives to sync before they are powered off. Reason: When I import the pool as root, after failure, I see some writes to the dis(s). Only a couple of seconds worth, that I didn't see before when it failed on reboot. Am I explaining myself very well?
More/better info. I have, what appears to be a workaround. Shutdown isn't the problem. Startup is. I have modified fuse-helper script: Added a sleep for 16 seconds. ... start() { sleep 15 [ -x $exec ] || (echo "$prog binary not present or executable" && exit 5) PID=`pidofproc $prog` ... 10 seconds allows for zfs to start, but not automount the drives. 15 works. Everytime, so far. Ran a dracut -f. Still ok. Errors showed up again after some updates ( including Systemd and fuse-helper ). Left it alone for few more updates, no change. Finally added the sleep 15 ( started with 30 ). Somewhere there is a serious timing problem. I'm not educated enough to find where/what.
Still valid?
CC: (none) => fri