On 15/10/2019 21:32, Curtis Vaughan wrote:
On 10/14/19 5:44 PM, Wols Lists wrote:
On 15/10/19 00:56, Curtis Vaughan wrote:
I have reason to believe one HD in a RAID1 is dying. But I'm trying to
understand what the logs and results of various commands are telling me.
Searching on the Internet is very confusing. BTW, this is for and Ubuntu
Server 18.04.2 LTS.
Ubuntu LTS ... hmmm. What does "mdadm --version" tell you?
mdadm - v4.1-rc1 - 2018-03-22
That's good. A lot of Ubuntu installations seem to have mdadm 3.4, and
that has known issues, shall we say.
It seems to me that the following information is telling me on device is
missing. It would seem to me that sda is gone.
Have you got any diagnostics for sda? Is it still in /dev? has it been
kicked from the system? Or just the raid?
It is still on the system, so I guess it's just kicked off the raid.
Take a look at
https://raid.wiki.kernel.org/index.php/Linux_Raid#When_Things_Go_Wrogn
Especially the "asking for help" page. It gives you loads of diagnostics
to run.
Ok, I already went through most of those tests, but still not sure what
they are telling me.
That's why the page tells you to post the output to this list ...
I'm particularly interested in what mdadm --examine and --detail tell me
about sda. But given that you think the drive is dud, does that really
matter any more?
However, there's a lot of reasons drives get kicked off, that may have
absolutely nothing to do with them being faulty. Are your drives
raid-friendly? Have you checked for the timeout problem? It's quite
possible that the drive is fine, and you've just had a glitch.
Basically, we need to know whether sda has died, or whether it's a
problem with raid (especially with older mdadm, like I suspect you may
have, the problem could lie there).
Anyhow, for example, I received an email:
A DegradedArray event had been detected on md device /dev/md0.
Do you have a spare drive to replace sda? If you haven't, it might be an
idea to get one - especially if you think sda might have failed. In that
case, fixing the raid should be pretty easy. So long as fixing it
doesn't tip sdb over the edge ...
The replacement drive is coming tomorrow. I'm certain now there's a
major issue
with the drive and will be replacing it.
What makes you think that?
My intent is to basically follow these instructions for replacing the
drive.
sudo mdadm --remove /dev/md1 /dev/sda2
sudo mdadm --remove /dev/md0 /dev/sda1
Remove the bad drive, install new drive
sudo mdadm --add /dev/md1 /dev/sda2
sudo mdadm --add /dev/md1 /dev/sda1
Would that be the correct approach?
Yup. Sounds good. The only thing that might make sense, especially if
you're getting a slightly bigger drive to replace sda, look at putting
dm-integrity between the partition and the raid. There's a good chance
it's not available to you because it's a new feature, but the idea is
that it checksums the writes. So if there's data corruption the raid no
longer wonders which drive is correct, but the corruption triggers a
read error and the raid will fix it. I can't give you any pointers
sorry, because I haven't played with it myself, but you should be able
to find some information here or elsewhere on it.
Finally could you tell me how to subscribe to this newsgroup through an NNTP client?
I was using it through the gmane server, which seems to have issues of whether it's
being continued or not. And although I can see some recent posts from last week,
there has been nothing new. I've been searching for a NNTP server, but can't find one.
Thanks!
Dunno how to subscribe using an nntp client because this is a mailing
list, but details are on the wiki home page. Click on the link to the
mailing list, and you can subscribe to the list there.
Cheers,
Wol
