[prev in list] [next in list] [prev in thread] [next in thread] 

List:       linux-lvm
Subject:    Re: [linux-lvm] RAID1 mirroring behavior when one disk is dead sometimes - why not just use random n
From:       Marc Weber <marco-oweber () gmx ! de>
Date:       2021-04-15 4:19:50
Message-ID: 9a08e0e2-2423-c7f2-df96-94caf53eb0b3 () gmx ! de
[Download RAW message or body]


> As far as I know mdadm works the same way. (At least it was working
> like this when this thread happened:
> https://www.spinics.net/lists/raid/msg36962.html)
> 
> You can use a similar tactic that was recommended there: Don't
> assemble a degraded RAID1 array without user interaction. You can do
> this by changing the activation mode to complete in lvm.conf.

But then the server doesn't start at all which might also be costly.


The thread says that clock i not reliable eventually due to faulty hardware etc.

*BUT* why isn't there a way to 'stamp' the disks once clock is synced with the \
internet ?


Let's compare two cases and two solutions.


solution 1:

write dates [date, known-devices] to the active disks

CASE: disk 1 appears sometimes. disk2 would have 'know-devices' disk2 only, so disk1 \
could be identified as troublesome.

action mirror disk2   to1


CASE: split brain: disk1 appears, reboot disk2 appears, reboot both appear

Dates would be different. -> no mount


Not using the date info at all, thus could be using random numbers which would be \
solution 2.



I don't have experience how often disks appear sometimes only.

To me it looks like having such strategy above which also could be implemented in \
user space (but why ?)

would work in more cases very well cause the server would boot more often than not.


Marc Weber


_______________________________________________
linux-lvm mailing list
linux-lvm@redhat.com
https://listman.redhat.com/mailman/listinfo/linux-lvm
read the LVM HOW-TO at http://tldp.org/HOWTO/LVM-HOWTO/


[prev in list] [next in list] [prev in thread] [next in thread]