Re: [Vchanger-users] Non-mounted media volume's marked as error...

[Marco G. <ga...@li...>]

Mandi! Josh Fisher via Vchanger-users
  In chel di` si favelave...

>> To me, RDX seems very like any USB disks; i don't know how to debug UDEV
>> event, but if you say me how, i can try to look at...

> I don't think there is anything to debug. The only difference was that 
> at the time when the udev script was invoked, the filesystem was 
> unmounted, yet the device node /dev/sdX was still there. Usually, with a 
> USB drive, the device node is delete by the OS and that is what triggers 
> udev. With RDX it seems to be different, maybe because the RDX device is 
> still plugged in and powered. Only the cartridge was ejected. How do you 
> usually eject? Do you umount the filesystem first, or just eject the 
> cartridge?

I umount the FS, and then i eject it. But there's no '/dev/rdx' nor
permanent device:

root@cnpve3:~# df
Filesystem                    1K-blocks       Used   Available Use% Mounted on
udev                           12220504          0    12220504   0% /dev
tmpfs                           2449644      42276     2407368   2% /run
rpool/ROOT/pve-1              450961536    2081408   448880128   1% /
tmpfs                          12248204      58136    12190068   1% /dev/shm
tmpfs                              5120          4        5116   1% /run/lock
tmpfs                          12248204          0    12248204   0% /sys/fs/cgroup
rpool                         448880256        128   448880128   1% /rpool
rpool/data                    448880256        128   448880128   1% /rpool/data
rpool/ROOT                    448880256        128   448880128   1% /rpool/ROOT
rpool-backup                52487099648 9174830208 43312269440  18% /rpool-backup
/dev/fuse                         30720         36       30684   1% /etc/pve
10.10.5.23:/rpool-backup/vz 52487100416 9174830080 43312270336  18% /mnt/pve/nfs-scratch
/dev/sdj1                    1921720580  607441416  1216587308  34% /mnt/vchanger/4ccfcae1-8a35-4800-96b6-9dce8718569b
tmpfs                           2449640          0     2449640   0% /run/user/0

root@cnpve3:~# ls -la /dev/rdx*
ls: cannot access '/dev/rdx*': No such file or directory
root@cnpve3:~# ls -la /dev/sdj*
brw-rw---- 1 root disk 8, 144 Mar 31 14:18 /dev/sdj
brw-rw---- 1 root disk 8, 145 Mar 31 14:18 /dev/sdj1
root@cnpve3:~# ls -la /dev/disk/by-id/ | grep sdj
lrwxrwxrwx 1 root root    9 Mar 31 14:18 usb-HP_RDX_003C75030237-0:0 -> ../../sdj
lrwxrwxrwx 1 root root   10 Mar 31 14:18 usb-HP_RDX_003C75030237-0:0-part1 -> ../../sdj1


if i umount and eject it, the main device remains, effectively:

root@cnpve3:~# ls -la /dev/rdx*
ls: cannot access '/dev/rdx*': No such file or directory
root@cnpve3:~# ls -la /dev/sdj*
brw-rw---- 1 root disk 8, 144 Apr  5 18:23 /dev/sdj
root@cnpve3:~# ls -la /dev/disk/by-id/ | grep sdj
lrwxrwxrwx 1 root root    9 Apr  5 18:23 usb-HP_RDX_003C75030237-0:0 -> ../../sdj

Interesting also the 'cfdisk' error:

root@cnpve3:~# cfdisk /dev/sdj
cfdisk: cannot open /dev/sdj: No medium found


So, effectively, a device remain.


>> 2) so, seems a Bacula bug? If yes, it worth a try to move to bacula list?
> Yes. I think so. When Bacula is looking for an available volume for a 
> job, it at some point tries to make a volume available by recycling old 
> volumes. If it is able to recycle a volume that the job can use, then it 
> sees an available volume. When it sees the available volume, it tries to 
> load it, even if that volume is not in a slot. When it fails to load, it 
> causes an i/o error (file not found) and that causes Bacula to mark the 
> volume in error. The bug is that Bacula should not ever try to load a 
> volume that is not in a slot when reading/writing to an autochanger device.

I'll try it. Thanks.

-- 
  E quindi vado avanti e non mi svesto,
  dei panni che son solito portare			(F. Guccini)