2

S.M.A.R.T. has found an unrecoverable read-error on one of my disks, but zpool status lists all disks as ONLINE (I.E. not DEGRADED).

  1. Do you know why might that be? I though ZFS would know of any errors as soon as anyone...
  2. Do I need to run a scrub in order for it to recheck the status of all disks?
  3. Can I have S.M.A.R.T. automatically report to ZFS somehow?
lindhe
  • 4,096
  • 3
  • 20
  • 34
  • What is the output of `zpool status -v`? It's possible zfs detected and repaired the error. It's also possible that zfs when attempted to write to a sector, the drive determined it couldn't, and has mapped the write somewhere else without letting zfs know. – mmusante Nov 29 '17 at 00:36

1 Answers1

2
  1. Do you know why might that be? I though ZFS would know of any errors as soon as anyone...
  2. Do I need to run a scrub in order for it to recheck the status of all disks?
  3. Can I have S.M.A.R.T. automatically report to ZFS somehow?
  1. No, it does not check all blocks all the time, it just makes sure that each written block can be accounted for (and restored, if redundancy is available) as soon as it is needed/accessed. Empty space is not checked at all (because you don't have valuable data there, so it would be a waste of time), and normal data is only checked when it is read (as write is append-only). As mmusante correctly said, you will only get error messages if the error is critical and can not be recovered from automatically (otherwise, you just see a notice and error counts in zpool status).
  2. Yes.

  3. It may be easier to just regularly (via cronjob) scrub the pool. Common recommended times are about once a month for enterprise-quality disks and once a week for consumer-level disks. Otherwise you could start a manual scrubbing with a script from smartmontools:

    Most of the time, you only need to place a script in /etc/smartmontools/run.d/. Whenever smartd wants to send a report, it will execute smart-runner and the latter will run your script.

    You have several variables available to your script (again, see the smartd manpage). These come from a test run:

    SMARTD_MAILER=/usr/share/smartmontools/smartd-runner
SMARTD_SUBJECT=SMART error (EmailTest) detected on host: XXXXX
SMARTD_ADDRESS=root
SMARTD_TFIRSTEPOCH=1267409738
SMARTD_FAILTYPE=EmailTest
SMARTD_TFIRST=Sun Feb 28 21:45:38 2010 VET
SMARTD_DEVICE=/dev/sda
SMARTD_DEVICETYPE=sat
SMARTD_DEVICESTRING=/dev/sda
SMARTD_FULLMESSAGE=This email was generated by the smartd daemon running on:
SMARTD_MESSAGE=TEST EMAIL from smartd for device: /dev/sda

    Your script also has a temporary copy of the report available as "$1". It will be deleted after you finish but the same content is written to /var/log/syslog.

    You then just need to map from the device name to your pool (you can parse zpool status).

user121391
  • 1,272
  • 7
  • 19
  • I think my main "issue" was that I had not run a scrub since the fault started to appear. I run scrubs once a month, but smart tests every day since they are much quicker. – lindhe Nov 30 '17 at 17:54