[
https://issues.apache.org/jira/browse/HDDS-1209?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16788898#comment-16788898
]
Nanda kumar commented on HDDS-1209:
-----------------------------------
Thanks [~avijayan] for working on this, overall the patch looks good to me.
We can replace the below logic
{code}
ContainerInfo containerInfo = null;
// Checking if excludedContainers is empty so that we don't loop around
// containerIDs for every allocate block.
if (CollectionUtils.isNotEmpty(excludedContainers)) {
NavigableSet<ContainerID> allowedContainerIDs = containerIDs
.parallelStream()
.filter(containerID -> !excludedContainers.contains(containerID))
.collect(Collectors.toCollection(TreeSet<ContainerID>::new));
containerInfo = containerStateManager.getMatchingContainer(
sizeRequired, owner, pipeline.getId(), allowedContainerIDs);
} else {
containerInfo = containerStateManager.getMatchingContainer(
sizeRequired, owner, pipeline.getId(), containerIDs);
}
{code}
with
{code}
containerIDs.removeAll(excludedContainers);
ContainerInfo containerInfo =
containerStateManager.getMatchingContainer(sizeRequired, owner,
pipeline.getId(), containerIDs);
{code}
Can you also add a test case to verify if we are creating a new container in
the selected pipeline if all the existing containers are in the excluded list?
> Fix the block allocation logic in SCM when client wants to exclude all
> available open containers in a chosen pipeline
> ---------------------------------------------------------------------------------------------------------------------
>
> Key: HDDS-1209
> URL: https://issues.apache.org/jira/browse/HDDS-1209
> Project: Hadoop Distributed Data Store
> Issue Type: Improvement
> Reporter: Shashikant Banerjee
> Assignee: Aravindan Vijayan
> Priority: Major
> Attachments: HDDS-1209-000.patch, HDDS-1209-001.patch
>
>
> As per [~nandakumar131] comments here in HDDS-726 ,
> https://issues.apache.org/jira/browse/HDDS-726?focusedCommentId=16779554&page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel#comment-16779554
> We may end up in case in SCM where all the open containers available in a
> chosen pipeline are there in the exclude list provided by client during
> allocate block call. In such case, blockMamager in SCM will run in an
> infinite loop. This needs to be fixed as a part of this jira.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
