Moving a server over to a RAID

[Mike A. Leonetti]

 (2010年08月22日 09:14), Luis Paulo wrote:
> On Sat, Aug 21, 2010 at 9:07 PM, Mike A. Leonetti
> <mikealeonetti at gmail.com> wrote:
>>  (2010年08月21日 15:27), Luis Paulo wrote:
> <snip>
>>> Let me just add that, with your new setup, you may create just one
>>> swap device (say /dev/md0) with /dev/sda2 and /dev/sdb2 in a mdadm
>>> raid0 array. Not that important as most systems rarely use it :)
>>>
>> Thanks mon. I think I read somewhere if you use the options
>> defaults,pri=1 for each swap you don't have to RAID them because the
>> kernel will automatically balance them or something like that.
> As you know, a raid0 swap will make disk access to swap faster, that
> was my idea.
>
> But as James said, if you want to be sure your server will keep
> running in a disk failure event, you should put the swap also in a
> raid1, right.
>
> On my server I just want to be able to restore it easy and fast, so I
> use raid1, but I put swap, /tmp and most of /var on raid0. (A disk
> failure is a disk failure, is a disk failure, is a ... :)
>
> BTW, a question to all, if you have a full raid1 (or 5) system, how do
> you know a disk has failed? I know you configure a mail address on
> mdadm.conf that reports it, ok, but you won't know it until you have
> read the mail. Do any of you have a way to get informed more "on the
> spot"?
> (Somehow this seems a stupid question, don't really know why)
>
> Thanks
> Luis
>
What I do is I have nagios monitor all servers with RAID monitored by
nagios by having NRPE and send e-mail alerts when /proc/mdstat reports
that a device is down (or shown as _ and not U).  The script I found
online was a Perl script that I can't seem to find anymore but that
really helps keep everything monitored.  Since nagios will keep polling
and depending on your settings keep sending alert e-mails you will
likely get the e-mail and be notified.