[sheepdog-users] Failing disk tests: disk not responding

[Valerio Pachera]

2013/10/17 Liu Yuan <namei.unix at gmail.com>:
> There is only one event will trigger auto unplug, that is EIO of the broken disk
> when client accesses it.

Ok.
I was wondering about the amount of time used by sheepdog to declare a
disk as failed (and auto unplug it).

I repeated the test and it takes about 10 minutes.
Note: this time I physically removed the disk instead of doing it by
software (echo 1 > /sys/block/sdc/device/delete).

Oct 18 16:44:24  ERROR [net 6677] do_read(220) connection is closed
(48 bytes left)
Oct 18 16:44:24  ERROR [net 6677] rx_work(583) failed to read a header
Oct 18 16:46:37  ERROR [io 6776] for_each_object_in_path(175) failed
to open /mnt/sheep/dsk03, Input/output error
Oct 18 16:47:00  ERROR [io 7061] for_each_object_in_path(175) failed
to open /mnt/sheep/dsk03, Input/output error
Oct 18 16:47:40  ERROR [io 6915] for_each_object_in_path(175) failed
to open /mnt/sheep/dsk03, Input/output error
Oct 18 16:48:19  ERROR [io 7062] for_each_object_in_path(175) failed
to open /mnt/sheep/dsk03, Input/output error
Oct 18 16:48:59  ERROR [io 7064] for_each_object_in_path(175) failed
to open /mnt/sheep/dsk03, Input/output error
Oct 18 16:50:07  ERROR [io 6911] for_each_object_in_path(175) failed
to open /mnt/sheep/dsk03, Input/output error
Oct 18 16:54:47  ERROR [io 7041] err_to_sderr(95) oid=83290e00000420,
Input/output error
Oct 18 16:54:47   INFO [main] md_remove_disk(316) /mnt/sheep/dsk03
from multi-disk array
Oct 18 16:54:47   INFO [main] recover_object_main(632) object
83290e00000420 is recovered (1/1082)
Oct 18 16:54:47   INFO [main] recover_object_main(632) object
5ddf88000000b1 is recovered (2/1082)

> Guest will not go to freeze when disks managed by sheepdog get broken, this is
> what a distributed system should provide as a bottom line.

Confirmed, this time I had a guest running on each of the 3 nodes and
they didn't stop nor shown errors.

Thank you.