Re: [ClusterLabs] newly created clone waits for off-lined node

2023-07-17 Thread lejeczek via Users




On 13/07/2023 17:33, Ken Gaillot wrote:

On Wed, 2023-07-12 at 21:08 +0200, lejeczek via Users wrote:

Hi guys.

I have a fresh new 'galera' clone and that one would not start &
cluster says:
...
INFO: Waiting on node  to report database status before Master
instances can start.
...

Is that only for newly created resources - which I guess it must be -
and if so then why?
Naturally, next question would be - how to make such resource start
in that very circumstance?

many thank, L.

That is part of the agent rather than Pacemaker. Looking at the agent
code, it's based on a node attribute the agent sets, so it is only
empty for newly created resources that haven't yet run on a node. I'm
not sure if there's a way around it. (Anyone else have experience with
that?)

any expert/devel would agree - this qualifies as a bug?
To add to my last message - mariadb starts a okey outside of 
the cluster but also if I manually change 
'safe_to_bootstrap' to '1' on any of the - in my case - two 
nodes, then cluster will also start the clone ok.
But then... disable(seems that Galera cluster gets shut down 
ok) & enable the clone and clone fails to start with those 
errors as earlier - waiting for off-lined node - and leaving 
the clone as 'Unpromoted'

___
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users

ClusterLabs home: https://www.clusterlabs.org/


Re: [ClusterLabs] newly created clone waits for off-lined node

2023-07-13 Thread lejeczek via Users




On 13/07/2023 17:33, Ken Gaillot wrote:

On Wed, 2023-07-12 at 21:08 +0200, lejeczek via Users wrote:

Hi guys.

I have a fresh new 'galera' clone and that one would not start &
cluster says:
...
INFO: Waiting on node  to report database status before Master
instances can start.
...

Is that only for newly created resources - which I guess it must be -
and if so then why?
Naturally, next question would be - how to make such resource start
in that very circumstance?

many thank, L.

That is part of the agent rather than Pacemaker. Looking at the agent
code, it's based on a node attribute the agent sets, so it is only
empty for newly created resources that haven't yet run on a node. I'm
not sure if there's a way around it. (Anyone else have experience with
that?)
But this is bit weird -- even with 'promoted-max' & 
'clone-max' to exclude (& on top of it ban the resource on 
the) off-lined node, clone would not start -- no?
Perhaps if not a an obvious bug, I'd think this could/should 
be improved, the code.
If anybody wanted to reproduce - have a "regular" galera 
clone okey, off-line a node, remove re/create the clone.

___
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users

ClusterLabs home: https://www.clusterlabs.org/


Re: [ClusterLabs] newly created clone waits for off-lined node

2023-07-13 Thread Ken Gaillot
On Wed, 2023-07-12 at 21:08 +0200, lejeczek via Users wrote:
> Hi guys.
> 
> I have a fresh new 'galera' clone and that one would not start &
> cluster says:
> ...
> INFO: Waiting on node  to report database status before Master
> instances can start.
> ...
> 
> Is that only for newly created resources - which I guess it must be -
> and if so then why?
> Naturally, next question would be - how to make such resource start
> in that very circumstance?
> 
> many thank, L.

That is part of the agent rather than Pacemaker. Looking at the agent
code, it's based on a node attribute the agent sets, so it is only
empty for newly created resources that haven't yet run on a node. I'm
not sure if there's a way around it. (Anyone else have experience with
that?)
-- 
Ken Gaillot 

___
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users

ClusterLabs home: https://www.clusterlabs.org/


[ClusterLabs] newly created clone waits for off-lined node

2023-07-12 Thread lejeczek via Users

Hi guys.

I have a fresh new 'galera' clone and that one would not 
start & cluster says:

...
INFO: Waiting on node  to report database status 
before Master instances can start.

...

Is that only for newly created resources - which I guess it 
must be - and if so then why?
Naturally, next question would be - how to make such 
resource start in that very circumstance?


many thank, L.

___
Manage your subscription:
https://lists.clusterlabs.org/mailman/listinfo/users

ClusterLabs home: https://www.clusterlabs.org/