Distribute may use send stream buffers inefficiently if some nodes stop early.

Description

If some nodes stop early, e.g. due to a choosen, distribute stops sending to those nodes. However, the buckets it had built up to that point that targetted those nodes were not being released and remained part of the overall capped send buffer. Distribute would therefore be working with the remainder of the capped send buffer for the rest of the distribute.

Ensure that the defunct send buckets are released as soon as possible.

Conclusion

None

Activity

Show:
Richard Chapman
updated the WorkflowSeptember 18, 2017 at 2:23 PM
NewTriagedWorkflow
AbbeyDoreWorkflow
Richard Chapman
updated the WorkflowNovember 26, 2015 at 8:50 AM
TriagedGitHubWorkflowWithAwaiting
NewTriagedWorkflow
Richard Chapman
updated the WorkflowNovember 13, 2015 at 11:21 AM
TriagedGitHubWorkflow
TriagedGitHubWorkflowWithAwaiting
Richard Chapman
changed the PriorityJuly 23, 2015 at 4:04 PM
Not specified
Major
Richard Chapman
updated the Issue TypeJuly 23, 2015 at 4:04 PM
Bug
Improvement
Gavin Halliday
changed the StatusMarch 18, 2015 at 1:35 PM
Merge Pending
Resolved
Gavin Halliday
updated the Fix versionsMarch 18, 2015 at 1:35 PM
None
5.4.0
Gavin Halliday
updated the ResolutionMarch 18, 2015 at 1:35 PM
None
Fixed
HPCC JiraBot
changed the StatusMarch 16, 2015 at 3:21 PM
Active
Merge Pending
HPCC JiraBot
updated the Pull Request URLMarch 16, 2015 at 3:21 PM
None
https://github.com/hpcc-systems/HPCC-Platform/pull/7121
Jacob Cobbett-Smith
created the IssueMarch 16, 2015 at 3:11 PM
Fixed
Pinned fields
Click on the next to a field label to start pinning.

Details

Components

Assignee

Reporter

Priority

Fix versions

Created March 16, 2015 at 3:11 PM
Updated July 23, 2015 at 4:04 PM
Resolved March 18, 2015 at 1:35 PM

Flag notifications