Does Slackware have automatic scrubbing of software RAID arrays implemented? I don't think it does, but it can't hurt to ask here and get a definitive answer.

Related question: how can one tell when the last scrubbing was performed on a particular array?

Reading your first sentence, I understood "scrubbing" to simply mean wiping with extreme prejudice. For that I would have used "dd if=/dev/zero of=raiddevice bs=5120 count=a_gazillion". But then your second sentence has confused me, as after a scrubbing you would essentially not have an array anymore. Can you be a bit more clear by what you mean by 'scrubbing", or describe the after-effects of "scrubbing".

Scrubbing of md arrays has to be done by a cron job, and Slackware has no such job configured by default (I'm not aware of any distribution that has).

As for your second question, it seems (I'm hoping someone will prove me wrong) there's no field in the md metadata that keeps track of scrubbing once the job is completed. In other words, you can't tell from the metadata when (if ever) an array was last scrubbed; you'll have to check the logs.

The term "Scrubbing" means verifying every sector on every RAID member. It's also called "verifying" by some vendors. AMD calls it "synchronization", which is misleading to say the least.

The point is to detect and possibly correct errors before they become critical. Given the size of modern hard drives and the increasing popularity of delta backup schemes, large parts of an array can sit idle for very long periods. Should a sector in such a part of an array suffer silent corruption ("bit rot"), the sysadmin will be in for a nasty surprise when another drive eventually fails and is replaced, and the rebuild job runs into the bad sector.

Scrubbing/verifying means that any bad sector will be detected and, since the array is still redundant, possibly fixed if the affected drive can perform a sector remapping.

4 members found this post helpful.

Thank you @Ser. To be honest, it is a rather poor name, as the connotation of "wiping clean" lurks. The other two you mention are far better. I would have have possibly preferred "repair", or even "preventative maintained".

True, but the same term, "scrubbing", is used to describe basically the same function on systems with ECC RAM. It may not be very descriptive of the process, but at least it's consistently used.

Not sure if you'll find this helpful but I wrote an rc.mdadm that has a "check" function. It could be set up to run on a cronjob if you wish.

It "Works for me(TM)" but could probably do with some wider testing.

http://slackware.com/~mrgoblin/files/rc.mdadm

1 members found this post helpful.

Thanks mRgOBLIN. I was thinking of writing something myself, but now that I see the awk/sed magic that you put into it, I realize I can't possibly come up with something better.

Just one question about your script: why did you choose to make it an always-running process, with a PID file etc? Are there any advantages of doing it this way, that are not immediately evident?

The monitor function should run all the time to, umm, monitor. That's what the pid is for.

The scrubbing stuff doesn't use the pid.

As Mr Cranium alluded, having the monitor process running is a good idea. If you have put your email (e.g. MAILADDR root@mydomain.tld) into the mdadm.conf (as suggested by the comments in that file) you will get an email alert with any failures or other events that may need your attention.

Just add a line like this to your root crontab:

You'll see log activity when it runs:

While the scan is running, you can use "cat /proc/mdstat" to monitor the progress.