Date: Thu, 10 Jun 2010 11:53:24 +0200 From: Alexander Leidinger <Alexander@Leidinger.net> To: jhell <jhell@dataix.net> Cc: fs@freebsd.org Subject: Re: Do we want a periodic script for a zfs scrub? Message-ID: <20100610115324.10161biomkjndvy8@webmail.leidinger.net> In-Reply-To: <4C0FB1DE.9080508@dataix.net> References: <20100609162627.11355zjzwnf7nj8k@webmail.leidinger.net> <4C0FAE2A.7050103@dataix.net> <4C0FB1DE.9080508@dataix.net>
next in thread | previous in thread | raw e-mail | index | archive | help
Quoting jhell <jhell@dataix.net> (from Wed, 09 Jun 2010 11:23:10 -0400):
> On 06/09/2010 11:07, jhell wrote:
>> On 06/09/2010 10:26, Alexander Leidinger wrote:
>>> Hi,
>>>
>>> I noticed that we do not have an automatism to scrub a ZFS pool
>>> periodically. Is there interest in something like this, or shall I keep
>>> it local?
>>>
>>> Here's the main part of the monthly periodic script I quickly created:
>>> ---snip---
>>> case "$monthly_scrub_zfs_enable" in
>>> [Yy][Ee][Ss])
>>> echo
>>> echo 'Scrubbing of zfs pools:'
>>>
>>> if [ -z "${monthly_scrub_zfs_pools}" ]; then
>>> monthly_scrub_zfs_pools="$(zpool list -H -o name)"
>>> fi
>>>
>>> for pool in ${monthly_scrub_zfs_pools}; do
>>> # successful only if there is at least one pool to scrub
>>> rc=0
>>>
>>> echo " starting scrubbing of pool '${pool}'"
>>> zpool scrub ${pool}
>>> echo " consult 'zpool status ${pool}' for the result"
>>> echo " or wait for the daily_status_zfs mail, if
>>> enabled"
>>> done
>>> ;;
>>> ---snip---
>>>
>>> Bye,
>>> Alexander.
>>>
>>
>> Please add a check to see if any resilerving is being done on the pool
>> that the scub is being executed on. (Just in case), I would hope that
>> the scrub would fail silently in this case.
>>
>> Please also check whether a scrub is already running on one of the pools
>> and if so & another pool exists start a background loop to wait for the
>> first scrub to finish or die silently.
>>
>> I had a scrub fully restart from calling scrub a second time after being
>> more than 50% complete, its frustrating.
>>
>>
>> Thanks!,
>>
>
> I should probably suggest one check that comes to mind.
>
> zpool history ${pool} | grep scrub | tail -1 |cut -f1 -d.
>
> Then compare the output with today's date to make sure today is >= 30
> days from the date of the last scrub.
>
> With the above this could be turned into a daily_zfs_scrub_enable with a
> default daily_zfs_scrub_threshold="30" and ensuring that if one check is
> missed it will not take another 30 days to run the check again.
Good idea! I even found a command line which does the calculation for
the number of days between "now" and the last run (not taking a leap
year into account, but an off-by-one day error here does not matter).
Bye,
Alexander.
--
"He's a businessman. I'll make him an offer he can't refuse."
-- Vito Corleone, "Chapter 1", page 39
http://www.Leidinger.net Alexander @ Leidinger.net: PGP ID = B0063FE7
http://www.FreeBSD.org netchild @ FreeBSD.org : PGP ID = 72077137
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20100610115324.10161biomkjndvy8>