[ClusterLabs] Antw: [EXT] Re: Merging partitioned two_node cluster?

[Ulrich Windl]

>>> "Nickle, Richard" <rnickle at holycross.edu> schrieb am 05.05.2020 um 18:30
in
Nachricht
<19182_1588696251_5EB194BA_19182_3_1_CAD6eeWcH=QWPsnoOChEbgrrt_m-8W+Je9yLvE=mDmG
_vJCAhw at mail.gmail.com>:
> So I tried an experiment.  I had tried switch over to 'udpu' unicast
> transport, but corosync threw an error starting up (which I did not drill
> down on yet.)
> 
> I went over to my test environment and did the same thing and it worked
> fine, the cluster worked and everything.
> 
> One thing that is different there is the test environment is on a network
> with CIDR /22, and the production network is on a CIDR /26.
> 
> So, for my bindnetaddr on my test network, I have the address of my VIP, or
> '192.168.193.113'.  That would be a network base of '192.168.192.0'.
> 
> But my production network is on '192.168.83.131', That would be a network
> base of '192.168.83.128'.
> 
> I have tried hardcoding the 'bindnetaddr' to the network base
> '192.168.83.128', but it still throws an error.  Perhaps it is 'zeroing'
> out the bindnetaddr least significant byte to make the network base?
> 
> Is it possible that the calculation of my base network in 'bindnetaddr'
> doesn't account for networks with CIDR mask bits greater than 24?  (which
> would have non-zero least significant bytes.)

I wonder: Does "corosync-cfgtool -s" exist for your environment? If so, what's
the output on each node?

> 
> Thanks,
> 
> Rick
> 
> 
> 
> 
> 
> On Tue, May 5, 2020 at 12:03 PM Andrei Borzenkov <arvidjaar at gmail.com>
> wrote:
> 
>> 05.05.2020 16:44, Nickle, Richard пишет:
>> > Thanks Honza and Andrei (and Strahil?  I might have missed a message in
>> the
>> > thread...)
>> >
>>
>> Yep, all messages from Strahil end up in spam folder.
>> _______________________________________________
>> Manage your subscription:
>> https://lists.clusterlabs.org/mailman/listinfo/users 
>>
>> ClusterLabs home: https://www.clusterlabs.org/ 
>>