RFC: Disk Monitoring

[Brian Sutherland]

On Tue, Mar 22, 2005 at 04:19:58AM -0500, ioliver wrote:
> I'm afraid I don't know about about debconf to help with the bulk of
> your RFC, but I do have some questions/observations.

It's become clearer that the debian maintainer is comfortable with
leaving debconf out of it if the defaults are good. So hopefully we can
leave out that level of complexity.

> 1) Is it the DEVICESCAN that's causing some hardware issues? The
> default could be to have this commented out, but no entries to exist
> for specific devices. This ties in to -

Yep, it is DEVICESCAN. But debian upstream is in the process of solving
this.  Probably with a script which will generate a safe smartd.conf.

> 2) Would the graphical application also be used to select which devices
> are to be monitored and when self tests are to be scheduled? If so, then
> all monitoring could be off by default until enabled in this graphical
> app.

Ideally I would like this to be something that does not need to be
configured. i.e. "just works". That way, a graphical notifier could be
written that most people would not know exists until one day...

> 3) Some attributes on drives change very frequently. Smartmontools can
> generate a lot of log noise unless you tell it to ignore these
> attributes. You need to avoid worrying/bothering people.

Yes, definitely we shouldn't bother people or make them worry that their
hard drive is failing when it is not.

I think there are three places where we can do some kind of filtering:
    * script runner routes messages to 2 different script directories,
      i.e. /etc/smartd/all.d and /etc/smartd/critical.d
    * default config of smartd to not generate this noise
    * the graphical monitoring tool only chooses some messages it
      receives to display.

I am not sure which one to use right now, but if someone wants to help
out with a comment, welcome!

> 4) Drive temperature is reputed to have a major effect on longevity.
> The graphical app could show this as a thermometer bar, with some
> rather red bits towards the top end!

Would be cool, but I don't think there need to be any changes to
smartmontools for this, you can get the temperature from smartctl. So if
someone wants to write a graphical app for this they can probably do that
right now.

> Don't let the lack of feedback bother you too much. I think that people
> hold back when things are being discussed in the abstract. Wait until
> the 1st version hits - then you'll get feedback! :-)

No worries, I also think that the developers might be a little busy this
close to release. Perhaps I will post again in a month or so.

-- 
Brian Sutherland

It's 10 minutes, 5 if you walk fast.