Stanislav Lukyanov commented on IGNITE-8133:

Ignite now stores the baseline topology history. If a node attempts to join the 
cluster, and that node's baseline history is different than the history of the 
cluster, the node will be rejected to protect the cluster from adding a node 
that has data modified when the cluster was segmented.
Currently this behavior described only in the Javadoc of an internal class 
org.apache.ignite.internal.processors.cluster.BaselineTopology, so there seem 
to be no documentation a user (not a developer) can read to find out about it.

> Baseline topology documentation improvement
> -------------------------------------------
>                 Key: IGNITE-8133
>                 URL: https://issues.apache.org/jira/browse/IGNITE-8133
>             Project: Ignite
>          Issue Type: Improvement
>          Components: documentation
>    Affects Versions: 2.4
>            Reporter: Stanislav Lukyanov
>            Assignee: Stanislav Lukyanov
>            Priority: Critical
>             Fix For: 2.5
> Baseline topology concept was added to Ignite in 2.4 by IEP-4. This changed 
> Ignite cluster behavior when persistence is enabled (first of all, activation 
> and rebalancing timings).
> It seems that the current documentation may be confusing.
> For example, the sentence
> {quote}Note that the baseline topology is not set when the cluster is started 
> for the first time; that's the only time when a manual intervention is 
> needed.{quote}
> may lead one to think that baseline topology is not used by default and needs 
> to be enabled only if one wants to use it.
> Also, the documentation describes the tools and commands that are used to 
> manage the baseline topology and activation, but doesn't give guidelines on 
> which nodes should be in the topology, when should it be changed, etc.
> The documentation should be enhanced to
> - give clear understanding that baseline topology always needs to be 
> considered as a part of the cluster architecture when persistence is enabled;
> - provide overview of the behavioral changes compared to AI 2.3 (use a 
> note/warning block for that to separate it from the main text?);
> - provide basic guidelines and suggestions of how one can start a new cluster 
> and manage it (when to activate/deactivate, when to change baseline topology, 
> what happens and what needs to be done when a node fails or joins, how to use 
> consistentId)

This message was sent by Atlassian JIRA

Reply via email to