[gentoo-user] Failure to start md at boot

[gentoo-user] Failure to start md at boot

[Peter Humphrey]

I've split this out from the previous thread because it was getting messy. I'd 
followed Rich Freeman's advice to specify arrays by UUID in mdadm.conf, and 
this is what happened:

On Friday 29 May 2015 01:10:52 I wrote:

> OK, so this is what I have at present. I haven't booted with it yet to test
> it - I'll do that in the morning:
> 
> DEVICE /dev/sd[abcde][123456789]
> ARRAY /dev/md1 UUID=ea156c7f:183ca28e:c44c77eb:7ee19756
> ARRAY /dev/md5 UUID=e7640378:966a5b3a:c44c77eb:7ee19756
> ARRAY /dev/md7 UUID=c2d056c4:9118021f:ad73c633:b38fa97c

Specifying the UUIDs hasn't helped. I still get failure to start /dev/md7 
during boot as often as not.

-- 
Rgds
Peter

Re: [gentoo-user] Failure to start md at boot

[Mick]

[-- Attachment #1: Type: Text/Plain, Size: 1347 bytes --]

On Sunday 31 May 2015 10:01:43 Peter Humphrey wrote:
> I've split this out from the previous thread because it was getting messy.
> I'd followed Rich Freeman's advice to specify arrays by UUID in
> mdadm.conf, and this is what happened:
> 
> On Friday 29 May 2015 01:10:52 I wrote:
> > OK, so this is what I have at present. I haven't booted with it yet to
> > test it - I'll do that in the morning:
> > 
> > DEVICE /dev/sd[abcde][123456789]
> > ARRAY /dev/md1 UUID=ea156c7f:183ca28e:c44c77eb:7ee19756
> > ARRAY /dev/md5 UUID=e7640378:966a5b3a:c44c77eb:7ee19756
> > ARRAY /dev/md7 UUID=c2d056c4:9118021f:ad73c633:b38fa97c
> 
> Specifying the UUIDs hasn't helped. I still get failure to start /dev/md7
> during boot as often as not.

Did you try my suggestion to (also) specify the metadata type to see if it 
makes a difference?

Something else to check:

Make sure that the /etc/fstab name for the RAID device and your 
/etc/mdadm.conf use the same name.

If you have changed the partitions on this RAID, or recreated it, make sure to 
run 'mdadm --zero-superblock', before you delete the partition from the 
partition table, or you could recover it when you recreate a partition if not 
zeroed.

Have a look at smartctl output to see if there is something wrong with any of 
the md7 disks.

-- 
Regards,
Mick

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

Re: [gentoo-user] Failure to start md at boot

[Peter Humphrey]

On Sunday 31 May 2015 13:06:58 Mick wrote:
> On Sunday 31 May 2015 10:01:43 Peter Humphrey wrote:
> > On Friday 29 May 2015 01:10:52 I wrote:
> > > OK, so this is what I have at present. I haven't booted with it yet to
> > > test it - I'll do that in the morning:
> > > 
> > > DEVICE /dev/sd[abcde][123456789]
> > > ARRAY /dev/md1 UUID=ea156c7f:183ca28e:c44c77eb:7ee19756
> > > ARRAY /dev/md5 UUID=e7640378:966a5b3a:c44c77eb:7ee19756
> > > ARRAY /dev/md7 UUID=c2d056c4:9118021f:ad73c633:b38fa97c
> > 
> > Specifying the UUIDs hasn't helped. I still get failure to start /dev/md7
> > during boot as often as not.
> 
> Did you try my suggestion to (also) specify the metadata type to see if it
> makes a difference?

Ah, no, I missed that. I'll try it and see how it goes.

> Something else to check:
> 
> Make sure that the /etc/fstab name for the RAID device and your
> /etc/mdadm.conf use the same name.

Yes, they're the same - but fstab isn't consulted until long after the VGs are 
started, is it?

> If you have changed the partitions on this RAID, or recreated it, make sure
> to run 'mdadm --zero-superblock', before you delete the partition from the
> partition table, or you could recover it when you recreate a partition if
> not zeroed.

Nope, haven't changed anything since creation.

> Have a look at smartctl output to see if there is something wrong with any
> of the md7 disks.

I was suspicious of the spinning disks I've just replaced with SSDs, because 
they were showing a few errors and not a lot of remaining life. Smartctl 
doesn't recognise the SSDs so I don't know how useful its output is, but it's 
not reporting anything anyway.

-- 
Rgds
Peter

Re: [gentoo-user] Failure to start md at boot

[Peter Humphrey]

On Sunday 31 May 2015 15:18:19 Peter Humphrey wrote:
> On Sunday 31 May 2015 13:06:58 Mick wrote:
> > On Sunday 31 May 2015 10:01:43 Peter Humphrey wrote:
> > > On Friday 29 May 2015 01:10:52 I wrote:
> > > > OK, so this is what I have at present. I haven't booted with it yet to
> > > > test it - I'll do that in the morning:
> > > > 
> > > > DEVICE /dev/sd[abcde][123456789]
> > > > ARRAY /dev/md1 UUID=ea156c7f:183ca28e:c44c77eb:7ee19756
> > > > ARRAY /dev/md5 UUID=e7640378:966a5b3a:c44c77eb:7ee19756
> > > > ARRAY /dev/md7 UUID=c2d056c4:9118021f:ad73c633:b38fa97c
> > > 
> > > Specifying the UUIDs hasn't helped. I still get failure to start
> > > /dev/md7
> > > during boot as often as not.
> > 
> > Did you try my suggestion to (also) specify the metadata type to see if it
> > makes a difference?
> 
> Ah, no, I missed that. I'll try it and see how it goes.

It hasn't helped.

I think I know what the problem is: namely, that I haven't discovered how to 
set up lvmetad. I get an error "lvmetad is running but disabled" but if I 
enable it in /etc/lvm/lvm.conf I get a different error. What should I be doing 
here? Google hasn't helped me.

In my own defence, I did mention this on Wednesday.

-- 
Rgds
Peter

Re: [gentoo-user] Failure to start md at boot

[Mick]

[-- Attachment #1: Type: Text/Plain, Size: 1598 bytes --]

On Sunday 31 May 2015 16:09:22 Peter Humphrey wrote:
> On Sunday 31 May 2015 15:18:19 Peter Humphrey wrote:
> > On Sunday 31 May 2015 13:06:58 Mick wrote:
> > > On Sunday 31 May 2015 10:01:43 Peter Humphrey wrote:
> > > > On Friday 29 May 2015 01:10:52 I wrote:
> > > > > OK, so this is what I have at present. I haven't booted with it yet
> > > > > to test it - I'll do that in the morning:
> > > > > 
> > > > > DEVICE /dev/sd[abcde][123456789]
> > > > > ARRAY /dev/md1 UUID=ea156c7f:183ca28e:c44c77eb:7ee19756
> > > > > ARRAY /dev/md5 UUID=e7640378:966a5b3a:c44c77eb:7ee19756
> > > > > ARRAY /dev/md7 UUID=c2d056c4:9118021f:ad73c633:b38fa97c
> > > > 
> > > > Specifying the UUIDs hasn't helped. I still get failure to start
> > > > /dev/md7
> > > > during boot as often as not.
> > > 
> > > Did you try my suggestion to (also) specify the metadata type to see if
> > > it makes a difference?
> > 
> > Ah, no, I missed that. I'll try it and see how it goes.
> 
> It hasn't helped.
> 
> I think I know what the problem is: namely, that I haven't discovered how
> to set up lvmetad. I get an error "lvmetad is running but disabled" but if
> I enable it in /etc/lvm/lvm.conf I get a different error. What should I be
> doing here? Google hasn't helped me.
> 
> In my own defence, I did mention this on Wednesday.

Sorry, but I don't really know much about LVM because I've never used it in 
anger.  I just (re)size my partitions as I need them to be.  However, I see 
this mentioned in the Wiki:

https://wiki.gentoo.org/wiki/Talk:LVM

HTH.
-- 
Regards,
Mick

[-- Attachment #2: This is a digitally signed message part. --]
[-- Type: application/pgp-signature, Size: 473 bytes --]

Re: [gentoo-user] Failure to start md at boot

[Peter Humphrey]

On Sunday 31 May 2015 16:34:19 Mick wrote:
> On Sunday 31 May 2015 16:09:22 Peter Humphrey wrote:
> > On Sunday 31 May 2015 15:18:19 Peter Humphrey wrote:
> > > On Sunday 31 May 2015 13:06:58 Mick wrote:
> > > > On Sunday 31 May 2015 10:01:43 Peter Humphrey wrote:
> > > > > On Friday 29 May 2015 01:10:52 I wrote:
> > > > > > OK, so this is what I have at present. I haven't booted with it
> > > > > > yet
> > > > > > to test it - I'll do that in the morning:
> > > > > > 
> > > > > > DEVICE /dev/sd[abcde][123456789]
> > > > > > ARRAY /dev/md1 UUID=ea156c7f:183ca28e:c44c77eb:7ee19756
> > > > > > ARRAY /dev/md5 UUID=e7640378:966a5b3a:c44c77eb:7ee19756
> > > > > > ARRAY /dev/md7 UUID=c2d056c4:9118021f:ad73c633:b38fa97c
> > > > > 
> > > > > Specifying the UUIDs hasn't helped. I still get failure to start
> > > > > /dev/md7
> > > > > during boot as often as not.
> > > > 
> > > > Did you try my suggestion to (also) specify the metadata type to see
> > > > if
> > > > it makes a difference?
> > > 
> > > Ah, no, I missed that. I'll try it and see how it goes.
> > 
> > It hasn't helped.
> > 
> > I think I know what the problem is: namely, that I haven't discovered how
> > to set up lvmetad. I get an error "lvmetad is running but disabled" but if
> > I enable it in /etc/lvm/lvm.conf I get a different error. What should I be
> > doing here? Google hasn't helped me.
> > 
> > In my own defence, I did mention this on Wednesday.
> 
> Sorry, but I don't really know much about LVM because I've never used it in
> anger.  I just (re)size my partitions as I need them to be.  However, I see
> this mentioned in the Wiki:
> 
> https://wiki.gentoo.org/wiki/Talk:LVM

Yes, I'd seen that, and several other sites that shed a little light.

One thing that does seem to have helped is that I've removed mdraid from the 
boot run level. Bug 521280 refers. It seems that mdraid must be in the boot 
run level during installation and the first few boots of the system, but should 
be removed again when the error "mdadm: No arrays found in config file or 
automatically" starts appearing at boot time.

-- 
Rgds
Peter

Re: [gentoo-user] Failure to start md at boot

[Peter Humphrey]

On Sunday 31 May 2015 10:01:43 I wrote:
> I've split this out from the previous thread because it was getting messy.
> I'd followed Rich Freeman's advice to specify arrays by UUID in mdadm.conf,
> and this is what happened:
> 
> On Friday 29 May 2015 01:10:52 I wrote:
> > OK, so this is what I have at present. I haven't booted with it yet to
> > test
> > it - I'll do that in the morning:
> > 
> > DEVICE /dev/sd[abcde][123456789]
> > ARRAY /dev/md1 UUID=ea156c7f:183ca28e:c44c77eb:7ee19756
> > ARRAY /dev/md5 UUID=e7640378:966a5b3a:c44c77eb:7ee19756
> > ARRAY /dev/md7 UUID=c2d056c4:9118021f:ad73c633:b38fa97c
> 
> Specifying the UUIDs hasn't helped. I still get failure to start /dev/md7
> during boot as often as not.

What I hope will be the final postscript:

I did eventually find the right setup, thus:

# cat /etc/mdadm.conf
# mdadm configuration file
#
DEVICE /dev/sd[ab][123456789]
ARRAY /dev/md1 metadata=0.90 UUID=0240695f:38fe6523:7bfe4778:da957cc1
ARRAY /dev/md5 metadata=0.90 UUID=a6a8a1bf:f7058f1d:7bfe4778:da957cc1
ARRAY /dev/md7 metadata=1.2 UUID=77f93f10:1ff22cae:81c5117d:5c25873e
ARRAY /dev/md8 metadata=0.90 UUID=ce921c64:3685e8c7:7bfe4778:da957cc1
ARRAY /dev/md9 metadata=1.2 UUID=0f74fff3:cdae1a64:0a44d949:4f1ad406

But I was still getting occasional failures of /dev/md7 to start - this 
contains all except the root and boot partitions of my everyday system, in 
LVM. It turns out that I had the wrong UUID specified in the file above, but 
nevertheless, almost always the system found the right devices to start. Now 
that I've corrected it I don't expect any more failures (that's what I meant 
when I said I hope this is the final postscript.)

In case anyone's interested, I've been installing a qt5 equivalent of my main 
system, using md8 as the root partition and md9 (plus LVM) for everything 
else. Oh, except /boot, which is md1 in common with the main system. My fstabs 
are getting long and complicated!

-- 
Rgds
Peter