HAST and CARP

[Freddie Cash]

On Sun, Jul 4, 2010 at 10:05 PM, hiroshi at soupacific.com
<hiroshi at soupacific.com> wrote:
>> I think you could make a configuration so when hostB (secondary) switches
>> to master it changes advskew to the value lower then on hostA.
>
>> In this way after hostA reboot it will have higher advskew and will be forced to do as a
>> secondary. Then after the nodes are synchronized you can switch to the
>> initial state restoring initial value of advskew on hostB.
>>
> NO ! Once split-brain mode appeared, CARP and HAST as BACKUP mode by changed
> whatever, advskew value etc. No way to get out slip-brain complain forever!
>
> Is there anybody having same problem ? Only me?

Once you are in a split-brain situation, you have to take manual steps
to repair.

Set one side as master.

Then run "hastctl create" on the other box, to reset all the hast
metadata on the devices, and initiate a new sync from the master.

Ideally, any automated scripts would handle all the possible error
conditions and checks, and prevent the systems from getting into the
split-brain situation in the first place.  :)  (Yeah, a lot easier
said than done.)

-- 
Freddie Cash
fjwcash at gmail.com