Github user felixcheung commented on the issue:
https://github.com/apache/spark/pull/18465
hmm, this is a fairly small but crucial change.
One point we were discussing, earlier, we are saying that we don't know if
the child is going away, or merely slow starting, so we explicitly send and
look for a exit code. I guess this condition is covered in the new change.
According to the doc
```
readChild and readChildren return a raw vector with a "pid" attribute if
data were available, integer vector of length one with the process ID if a
child terminated
```
So I guess it is safer to act only if we get an integer (which, according
to above, the child terminated), but to clarify
- isn't being "terminated" enough? that we have to pskill it again?
- the original issue with leaking was thought to be related to child
process getting stuck and *not* terminated properly? would that manifest again
under this new behavior? in other words, would it get into a state where
integer is never returned to the master?
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
