[ClusterLabs] Antw: Re: Random failure with clone of IPaddr2

[alian at amisw.com]

>>> I use same corosync / pacemaker on three host, but:
>>> Host A & B have same kernel, C is different:
>>> A & B : 3.7.10-1.45-desktop
>>> C: 4.1.15-8-default
>>
>> I don't know why you do that, but I'd either put C into standby, or put
>> A
>> nad B into standby and upgrade them one by one to the versions of C
>> (undo
>> standby after that). Your configuration makes troubleshooting a
>> challenge
>> (for you and others)!
>
> For pleasure ... as I like rebuild rpm from source ... Without joke, I
> start use cororysnc 3 years ago on A & B, but I need to upgrade these
> hosts, without losing connection or load-balancing. So the only way I
> found is to add a third host, to upgrade the first one. But when I get my
> third host this isn't the old os, so I rebuild rpm from source for A & B,
> and now I hit this wall. I can't put all services on C hosts. So no it
> isn't really for pleasure ;-)
>
> And after this night, I think the problem didn't come from corosync /
> pacemaker but from ipt_clusterip module ... I know this is isn't a good
> idea this config of different, but I try to find all the way to get out of
> this.
>

I check other configuration I have.
This work fine with :
- 3.11.10-25-default / 3.11.10-29-default
- 3.7.10-1.1-desktop / 3.11.10-29-default
- 3.7.10-1.16-desktop / 3.11.10-21-default

But so don't work with:
- 3.7.10-1.45-desktop / 4.1.15-8-default

I totally understand the problem didn't come from corosync / pacemaker
stuff, but from ip kernel 4.x configuration, around ipt_clusterip just
like this kernel didn't choose same "modulo" to answer or not to a packet.
(I enable debug trace of the module).

The upgrade process in production environment is not a easy stuff for sure
...