Github user steveloughran commented on the pull request:
https://github.com/apache/spark/pull/6935#issuecomment-161240157
the reason the modtime doesn't change is the log file is kept open; the mod
time is set when the output stream is created and not updated as new data is
added. The only viable probe for new data is going to be looking for new data.
This probe may need to handle incomplete data; the event log listener does
do a {{flush()}} on major events âthat is not enough to guarantee that
changes are visible across HDFS readers (more precisely, there are no
guarantees if/when writes to a file are picked up by existing input streams;
they may get an out of date view of the data, or an inconsistent one). Because
the input stream is re-opened on the playback, the risk is low âbut is it
there? If so, the risk has always existed when loading an incomplete app, it's
just with one load per run (ignoring cache evictions), there's significantly
fewer loads of incomplete apps than if it can be triggered by GETs on a loaded
app UI.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
