Andrew Beekhof writes:
> I believe things are much improved with more recent releases.
> Hence the recommendation of 1.4.x
Yes, we switched to corosync and the problem doesn't occur again!
All tests are passed now.
>>But really: you need to fix your "unsteady" network,
>>and probably should impl
On Wed, Mar 21, 2012 at 8:46 AM, Lars Ellenberg
wrote:
> On Sun, Mar 18, 2012 at 11:53:11PM +, Roman Sidler wrote:
>> Hi all
>> No, we have no crossover cable between the nodes.
>> The 2 nodes are linked by a switched network, and this works really fine
>> except
>> the mentioned case.
>>
>>
On Sun, Mar 18, 2012 at 11:53:11PM +, Roman Sidler wrote:
> Hi all
> No, we have no crossover cable between the nodes.
> The 2 nodes are linked by a switched network, and this works really fine
> except
> the mentioned case.
>
> It's rather easy to reproduce.
> 1. Activate a 2-node cluster
Andrew Beekhof writes:
>
> On Mon, Mar 19, 2012 at 10:53 AM, Roman Sidler
> wrote:
> The CCM (part of heartbeat) is known not to be particularly awesome in
> such scenarios and it appears to be the case again here:
>
> Mar 16 01:35:31 oan1 crmd: [7601]: info: ccm_event_detail: NEW MEMBERSHIP:
On Mon, Mar 19, 2012 at 10:53 AM, Roman Sidler
wrote:
> Hi all
> No, we have no crossover cable between the nodes.
> The 2 nodes are linked by a switched network, and this works really fine
> except
> the mentioned case.
>
> It's rather easy to reproduce.
> 1. Activate a 2-node cluster
> 2. disco
Hi all
No, we have no crossover cable between the nodes.
The 2 nodes are linked by a switched network, and this works really fine except
the mentioned case.
It's rather easy to reproduce.
1. Activate a 2-node cluster
2. disconnect network connection (e.g by disconnect network adapter on VM)
3. w
