[
https://issues.apache.org/jira/browse/HDDS-8503?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Siddhant Sangwan updated HDDS-8503:
-----------------------------------
Description:
The java doc says of chooseNodes says that the nodes it will choose should be
on racks other than excluded nodes:
{code}
/**
* Choose a batch of datanodes on different rack than excludedNodes or
* chosenNodes.
* TODO HDDS-7226: Update Implementation to accomodate for already used
* nodes to conform to existing placement policy.
*
* @param excludedNodes - list of the datanodes to excluded. Can be null.
* @param chosenNodes - list of nodes already chosen. These nodes should also
* be excluded. Cannot be null.
* @param favoredNodes - list of favoredNodes. It's a hint. Whether the nodes
* are chosen depends on whether they meet the constrains.
* Can be null.
* @param usedNodes - list of the nodes that are already used.
* @param favorIndex - the node index of favoredNodes which is not chosen yet.
* @param nodesRequired - number of datanodes required.
* @param mapSizeRequired - size required for the container, Ratis metadata.
* @return List of chosen datanodes.
* @throws SCMException SCMException
*/
{code}
But the code looks like it's picking the same rack in one of the if-else
branches:
{code}
if (favoredNode != null && networkTopology.isSameParent(
excludedNodeList.get(excludedNodeList.size() - 1), favoredNode)) {
chosenNode = favoredNode;
favorIndex++;
}
{code}
Should probably be {{!networkTopology.isSameParent(...)}}.
We need to check if this is a bug or not.
was:
The java doc says of chooseNodes says that the nodes it will choose should be
on racks other than excluded nodes:
{code}
/**
* Choose a batch of datanodes on different rack than excludedNodes or
* chosenNodes.
* TODO HDDS-7226: Update Implementation to accomodate for already used
* nodes to conform to existing placement policy.
*
* @param excludedNodes - list of the datanodes to excluded. Can be null.
* @param chosenNodes - list of nodes already chosen. These nodes should also
* be excluded. Cannot be null.
* @param favoredNodes - list of favoredNodes. It's a hint. Whether the nodes
* are chosen depends on whether they meet the constrains.
* Can be null.
* @param usedNodes - list of the nodes that are already used.
* @param favorIndex - the node index of favoredNodes which is not chosen yet.
* @param nodesRequired - number of datanodes required.
* @param mapSizeRequired - size required for the container, Ratis metadata.
* @return List of chosen datanodes.
* @throws SCMException SCMException
*/
{code}
But the code looks like it's picking the same rack in one of the if-else
branches:
{code}
if (favoredNode != null && networkTopology.isSameParent(
excludedNodeList.get(excludedNodeList.size() - 1), favoredNode)) {
chosenNode = favoredNode;
favorIndex++;
}
{code}
Should probably be {{!networkTopology.isSameParent(...)}}.
> Possible bug when excluding rack in SCMContainerPlacementRackAware#chooseNodes
> ------------------------------------------------------------------------------
>
> Key: HDDS-8503
> URL: https://issues.apache.org/jira/browse/HDDS-8503
> Project: Apache Ozone
> Issue Type: Bug
> Components: SCM
> Affects Versions: 1.3.0
> Reporter: Siddhant Sangwan
> Assignee: Siddhant Sangwan
> Priority: Major
>
> The java doc says of chooseNodes says that the nodes it will choose should be
> on racks other than excluded nodes:
> {code}
> /**
> * Choose a batch of datanodes on different rack than excludedNodes or
> * chosenNodes.
> * TODO HDDS-7226: Update Implementation to accomodate for already used
> * nodes to conform to existing placement policy.
> *
> * @param excludedNodes - list of the datanodes to excluded. Can be null.
> * @param chosenNodes - list of nodes already chosen. These nodes should
> also
> * be excluded. Cannot be null.
> * @param favoredNodes - list of favoredNodes. It's a hint. Whether the
> nodes
> * are chosen depends on whether they meet the
> constrains.
> * Can be null.
> * @param usedNodes - list of the nodes that are already used.
> * @param favorIndex - the node index of favoredNodes which is not chosen
> yet.
> * @param nodesRequired - number of datanodes required.
> * @param mapSizeRequired - size required for the container, Ratis metadata.
> * @return List of chosen datanodes.
> * @throws SCMException SCMException
> */
> {code}
> But the code looks like it's picking the same rack in one of the if-else
> branches:
> {code}
> if (favoredNode != null && networkTopology.isSameParent(
> excludedNodeList.get(excludedNodeList.size() - 1), favoredNode)) {
> chosenNode = favoredNode;
> favorIndex++;
> }
> {code}
> Should probably be {{!networkTopology.isSameParent(...)}}.
> We need to check if this is a bug or not.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
