Github user dragos commented on the pull request:
https://github.com/apache/spark/pull/11207#issuecomment-184619429
@bbossy thanks for picking this up!
I have a problem with the bandwidth this design implies. For instance, my
state.json is 200KB (a cluster of 1 master and 2 nodes, with virtually no
frameworks running). This is most likely a very low minimum. Multiply this by
the number of nodes (let's say we want to scale to 10.000 nodes).
```
200KB * 10,000 nodes = 2GB
```
2GB of traffic for each check seems pretty bad.
I think it'd be better (and probably simpler) to increase the timeout, or
send a heartbeat message (whatever akka remote did before).
@tnachen @mgummelt what do you think?
---
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]
