[DRBD-user] fail-over using duplicate ROOT file system

Arnold Krille arnold at arnoldarts.de
Sun Jul 17 23:02:32 CEST 2011

Note: "permalinks" may not be as permanent as we would like,
direct links of old sources may well be a few messages off.


Hi,

you can work around the issue with the persistent-net-udev-rules by just 
deactivating these persistency rules...

But your setup is rather complicated and error-prone compared to a real 
failover-cluster with heartbeat/corosync. And syncing root-images with rsync 
will give you a state from while its running. At worst this won't boot 
correctly at all.
Whereas with a failover-cluster you get a running system with almost no 
interruption and no sysadmin interaction.

Have fun,

Arnold

On Sunday 17 July 2011 21:09:16 Ian! D. Allen wrote:
> Is this dual-ROOT scheme below workable?
> 
> Let me explain using Machine A and Machine B:
> 
> Machine A
>  - has a stock, ext4, bootable ROOT file system, non-DRBD
>  - has a second ROOT partition that is an RSYNC copy of Machine B
>  - has an ext4 data-only partition, DRBD primary, that uses
>    Machine B as DRBD secondary
> 
> Machine B
>  - has a stock, ext4, bootable ROOT file system, non-DRBD
>  - has a second ROOT partition that is an RSYNC copy of Machine A
>  - has a (hidden) data-only partition that is DRBD secondary, with
>    Machine A as DRBD primary.  (Hidden, because you can only mount ext4
>    on one machine, not two.)
> 
> Machine A does an RSYNC of its ROOT partition to the corresponding spare
> ROOT partition on Machine B regularly, and vice-versa.
> 
> When Machine A fails, reboot the Machine B hardware (DRBD secondary) using
> the RSYNC copy of Machine A's ROOT partition.  This brings the machine
> up solo as if it were Machine A, accessing the local DRBD partition as
> if it were Machine A (DRBD primary).  (Perhaps some fsck will be needed
> to make the ext4 on the DRBD usable?)  Is this doable?  (I know it works,
> since I've tried it, but I'm wondering if I'm missing anything.)
> 
> Work continues using Machine B's hardware, booted to be Machine A
> (DRBD primary).  This is now the new Machine A (DRBD primary), running
> without any Machine B (DRBD secondary).
> 
> Repair the failed old Machine A.  Boot the old Machine A hardware using
> the copy of Machine B's ROOT partition.  This brings the machine up
> as if it were Machine B, partnering with the new Machine A to provide
> secondary DRBD for the data-only partition.  This is now the new Machine B
> (DRBD secondary).
> 
> It it okay to bring up Machine B's secondary DRBD as primary, simply by
> using the copy of the ROOT from Machine A?  It it okay to make Machine
> A behave as Machine B, simply by using a copy of Machine B's ROOT?
> 
> One "gotcha" I've found is that booting Machine A's ROOT on Machine
> B's hardware leads to mis-numbering of the Machine B network
> cards as eth2 and eth3 due to Machine A entries already existing in
> /etc/udev/rules.d/70-persistent-net.rules.  I'll have to tweak those at
> boot time, or perhaps I can create a merged file that works correctly
> on both machines.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 198 bytes
Desc: This is a digitally signed message part.
URL: <http://lists.linbit.com/pipermail/drbd-user/attachments/20110717/da28e3ee/attachment.pgp>


More information about the drbd-user mailing list