vinum volumes crash on reboot

[aarong]

On Saturday, October 18, 2003, at 04:29  PM, Danny Pansters wrote:

> On Saturday 18 October 2003 07:48, aarong wrote:
>> I have a FreeBSD 4.8-RELEASE system with two identical 80GB IDE
>> drives. Vinum knows /dev/ad0 as "alpha" and "beta" is /dev/ad3. The
>> boxes root slice is a Vinum volume and everything on alpha runs
>> flawlessly. However for the past week I've tried to mirror alpha onto
>> beta without success. Whether adding plexes, or just creating a
>> simple 512MB volume on drive beta, Vinum will state beta as
>
> You don't create a volume on a drive. You use drives/partitions to make
> up subdisks which make up plexes (representing a pile of data) and with
> one or more plexes you make up volumes. And then you can mount a volume
> as a vinum partition.
>

Right.

>> "referenced" and, in the case below, plexes as faulty after
>
> "referenced" sounds like "was already used". Any disklabel output
> perhaps and fstab?

Referenced means the drive is referred too in Vinum's configuration but 
not in use; the state is neither down nor up, nor faulty, it's just 
sort of there. I'm trying to figure out why its only referenced and not 
used.

>
>> rebooting. What I'm having difficulty figuring out is why Vinum has a
>> problem, as none of the logs say why.
>>
>> # cat /var/log/vinum_history
>> ...
>> 19 Oct 2003 09:53:17.040730 *** vinum started ***
>> 19 Oct 2003 09:53:17.667679 list
>> 19 Oct 2003 09:53:21.147581 dumpconfig
>> 19 Oct 2003 09:53:32.324243 saveconfig
>> 19 Oct 2003 09:53:35.926307 create /vinum.mirror
>
> What's that? you mount a mirror onto /, /var, and so on. Your old
> ("single plex") vinum volumes have to be unmounted to mount the newly
> made mirror. I have the feeling you don't quite get the terminology.
> You may need to go into SU mode at the console.

I wasn't aware the volumes had to be umounted to add additional plexes 
to them. However this point is moot since I cannot create *any* plex on 
the second drive, "beta". Creating a "test" volume with a single plex 
and subdisk both on drive "beta" exhibit the exact same behavior: 
everything is fine until the system is rebooted and the volume crashes. 
Why?

>
>> drive beta device /dev/ad3s1h
>>   plex name swap.p1 org concat volume swap
>>    sd len 2096871s       driveoffset 265s drive beta
>>   plex name root.p1 org concat volume root
>>    sd len 256000s        driveoffset 2097136s drive beta
>>   plex name var.p1 org concat volume var
>>    sd len 8388608s       driveoffset 2353136s drive beta
>>   plex name usr.p1 org concat volume usr
>>    sd len 145554562s     driveoffset 10741744s drive beta
>
> This is identical to what your first disk has?

Completely identical. I can start the addition plexes without a 
problem, and they run perfectly. Only on reboot does everything come 
crashing down (excuse the pun ;-)

>
> Why are you saving the same config over again? This might be the 
> problem
> having a stale config. Vinum config goes on your disk. You put it up
> once (perhaps through the vinum prompt which is designed for it) then
> leave it.
>

Only in this case did I use saveconfig each time. You can see I really 
don't understand Vinum's behavior all that well; I was just trying 
something hoping it would work. In either case the saved configuration 
was fine, according to dumpconfig's output.

>> # cat /var/log/messages | grep vinum
>> ...
>> Oct 19 09:49:33  /kernel: Preloaded elf module "vinum.ko" at
>> 0xc05e209c. Oct 19 09:49:33  /kernel: Mounting root from
>> ufs:/dev/vinum/root Oct 19 09:53:36  /kernel: vinum: drive beta is up
>> Oct 19 09:53:36  /kernel: vinum: swap.p1 is faulty
>> Oct 19 09:53:36  /kernel: vinum: root.p1 is faulty
>> Oct 19 09:53:36  /kernel: vinum: var.p1 is faulty
>> Oct 19 09:53:36  /kernel: vinum: usr.p1 is faulty
>> Oct 19 09:59:31  /kernel: vinum: var.p1.s0 is up by force
>> Oct 19 09:59:31  /kernel: vinum: var.p1 is up
>> Oct 19 09:59:31  /kernel: vinum: var.p1.s0 is up
>> Oct 19 10:05:12  /kernel: vinum: swap.p1.s0 is up by force
>> Oct 19 10:05:12  /kernel: vinum: swap.p1 is up
>> Oct 19 10:05:12  /kernel: vinum: swap.p1.s0 is up
>
> This is synching with p0, which BTW is probably disk1 and the data
> contained in p0 corresponds with what's on disk0.''

No, this is on bootup. All *.p1 plexes were fine and running until the 
machine restarted. I'm just trying to figure out why.

> HTH

A little bit, thanks.


Regards,
-aarong

>
> Dan
>