[
https://issues.apache.org/jira/browse/HDDS-5804?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aravindan Vijayan updated HDDS-5804:
------------------------------------
Description:
>From the Javadoc of the class
{code}
* Container placement policy that randomly choose datanodes with remaining
* space to satisfy the size constraints.
* <p>
* The Algorithm is as follows, Pick 2 random nodes from a given pool of nodes
* and then pick the node which lower utilization. This leads to a higher
* probability of nodes with lower utilization to be picked.
* <p>
{code}
In SCMContainerPlacementCapacity#chooseNode, we choose the node with a higher
utilization % (scmUsed / capacity). This is used by the replication manager to
pick a node to replicate an under-replicated container to. This means that if
configured, this policy will pick the more used nodes, thereby creating the
opposite effect.
was:In SCMContainerPlacementCapacity#chooseNode, we choose the node with a
higher utilization % (scmUsed / capacity). This is used by the replication
manager to pick a node to replicate an under-replicated container to. This
means that if configured, this policy will pick the more used nodes, thereby
creating the opposite effect.
> SCMContainerPlacementCapacity is working incorrectly
> ----------------------------------------------------
>
> Key: HDDS-5804
> URL: https://issues.apache.org/jira/browse/HDDS-5804
> Project: Apache Ozone
> Issue Type: Bug
> Components: SCM
> Affects Versions: 1.2.0
> Reporter: Aravindan Vijayan
> Assignee: Aravindan Vijayan
> Priority: Major
>
> From the Javadoc of the class
> {code}
> * Container placement policy that randomly choose datanodes with remaining
> * space to satisfy the size constraints.
> * <p>
> * The Algorithm is as follows, Pick 2 random nodes from a given pool of nodes
> * and then pick the node which lower utilization. This leads to a higher
> * probability of nodes with lower utilization to be picked.
> * <p>
> {code}
> In SCMContainerPlacementCapacity#chooseNode, we choose the node with a higher
> utilization % (scmUsed / capacity). This is used by the replication manager
> to pick a node to replicate an under-replicated container to. This means that
> if configured, this policy will pick the more used nodes, thereby creating
> the opposite effect.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
