You are viewing a plain text version of this content. The canonical link for it is here.
Posted to user@ignite.apache.org by Raymond Wilson <ra...@trimble.com> on 2020/07/01 08:26:07 UTC

CusterGroup topology unexpectedly empty

I've been researching a problem where we see a ClusterGroup topology
becoming unexpectedly empty. I have not found anything that is similar to
my case outlined below:

We have some code using the C# client on Ignite 2.8.0 like this:

var _group = _ignite?.GetCluster()?.ForAttribute(_roleAttribute, "True");

There are nodes in the grid (deployed on AWS EKS) that have a local
attribute with the  _roleAttribute name set to the value "True", yet this
command sometimes decides that there are no matching nodes. Once it does
so, it does not seem to change requiring a restart of the grid.

In this instance, the _ignite reference may be a long lived reference to
the object returned by Ignition.GetIgnite("GridName") and cached for future
reference. Is this a safe practice, or can the underlying Ingite object
change leaving a long lived Ignite reference containing stale state
information about the grid?

Thanks,
Raymond.
.

-- 
<http://www.trimble.com/>
Raymond Wilson
Solution Architect, Civil Construction Software Systems (CCSS)
11 Birmingham Drive | Christchurch, New Zealand
+64-21-2013317 Mobile
raymond_wilson@trimble.com

<https://worksos.trimble.com/?utm_source=Trimble&utm_medium=emailsign&utm_campaign=Launch>

Re: CusterGroup topology unexpectedly empty

Posted by akurbanov <an...@gmail.com>.
Using ignite reference is completely fine.

What is the exact exception being thrown in this case? I would expect to see
this if the node is out of topology, what does the logs tell? Do other nodes
see this one in the topology?

Please check if there are NODE_FAILED events for the node on other topology
members, where the exception is observed.

Best regards,
Anton



--
Sent from: http://apache-ignite-users.70518.x6.nabble.com/

Re: CusterGroup topology unexpectedly empty

Posted by Ilya Kasnacheev <il...@gmail.com>.
Hello!

Please note that ClusterGroups usually do not recompute. This means, if you
had two nodes with _roleAttribute "True", and they all went away, the group
is now empty, even as two new nodes has joined with the same attribute.

You need to re-create ClusterGroup in order to refresh it.

Please check https://issues.apache.org/jira/browse/IGNITE-13050

Regards,
-- 
Ilya Kasnacheev


ср, 1 июл. 2020 г. в 11:26, Raymond Wilson <ra...@trimble.com>:

> I've been researching a problem where we see a ClusterGroup topology
> becoming unexpectedly empty. I have not found anything that is similar to
> my case outlined below:
>
> We have some code using the C# client on Ignite 2.8.0 like this:
>
> var _group = _ignite?.GetCluster()?.ForAttribute(_roleAttribute, "True");
>
> There are nodes in the grid (deployed on AWS EKS) that have a local
> attribute with the  _roleAttribute name set to the value "True", yet this
> command sometimes decides that there are no matching nodes. Once it does
> so, it does not seem to change requiring a restart of the grid.
>
> In this instance, the _ignite reference may be a long lived reference to
> the object returned by Ignition.GetIgnite("GridName") and cached for future
> reference. Is this a safe practice, or can the underlying Ingite object
> change leaving a long lived Ignite reference containing stale state
> information about the grid?
>
> Thanks,
> Raymond.
> .
>
> --
> <http://www.trimble.com/>
> Raymond Wilson
> Solution Architect, Civil Construction Software Systems (CCSS)
> 11 Birmingham Drive | Christchurch, New Zealand
> +64-21-2013317 Mobile
> raymond_wilson@trimble.com
>
>
> <https://worksos.trimble.com/?utm_source=Trimble&utm_medium=emailsign&utm_campaign=Launch>
>