Skip to:

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

Thor slave hang in countproject

Description

Very intermittent hang with countproject activity with multiple slaves
regression test mergededup.ecl can reproduce it if run many times

Conclusion

None

Activity

Show:

Jacob Cobbett-Smith August 23, 2016 at 4:32 PM

To happen, the countproject on a slave, must stop() before nextRow is ever called.
Which could happen if the downstream activities stop quickly, e.g. via a CHOOSEN.
e.g. if 1st slaves hit the CHOOSEN limit, other will quickly be told to stop().

In that case, the semaphore in COUNTPROJECT::onInputFinished will not have been signalled (normally signalled by 1st call to nextRow())

Meaning that the countproject::stop() when called will deadlock, since nextRow() will now never call it and the call to signalNext in countproject::stop() is after the PARENT::stop() that's stuck in the lookahead waiting in onInputFinished on the semaphore.

Mark Kelly August 22, 2016 at 2:19 PM

closeing to reissue similar fix for 6.0.6

Mark Kelly July 15, 2016 at 6:28 PM

Basically thread stuck in join because stop called before semaphore post

Mark Kelly July 15, 2016 at 6:28 PM

Should this go into 6.0.4 ?

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

Details

Components

Assignee

Reporter

Priority

Fix versions

Pull Request URL

https://github.com/hpcc-systems/HPCC-Platform/pull/9031
https://github.com/hpcc-systems/HPCC-Platform/pull/9031

Created July 15, 2016 at 6:22 PM
Updated August 24, 2016 at 12:57 PM
Resolved August 24, 2016 at 12:57 PM

Flag notifications