Skip to:

  1. Skip to Jira Navigation
  2. Skip to Side Navigation
  3. Skip to Main Content

Potential Smart Join deadlock if stopping LHS early

Description

If the LHS of a smart join has failed over to a local hash join and is stopped on a slave, before others have finished the hash distribute of the LHS a stall can occur.
The LHS stream (of the distributor) needs to be stopped before the distributor is closed.

Conclusion

None

Activity

Show:

Jacob Cobbett-Smith November 11, 2019 at 1:35 PM

The workaround if this issue is being hit in builds prior to this fix, would be to switch from using a SMART join, to either a regular HASH join, or a standard JOIN.
NB: because it only happens when it has already ran out of memory (and failed over to a local hash join), so coding specifically as a HASH join is no worse performance wise.

Jacob Cobbett-Smith November 1, 2019 at 3:07 PM

- FYI

Fixed
Pinned fields
Click on the next to a field label to start pinning.

Details

Components

Assignee

Reporter

Priority

Compatibility

Point

Fix versions

Pull Request URL

https://github.com/hpcc-systems/HPCC-Platform/pull/13117

Created October 31, 2019 at 5:10 PM
Updated November 11, 2019 at 1:35 PM
Resolved November 1, 2019 at 8:54 AM