[
https://issues.apache.org/jira/browse/BEAM-7520?focusedWorklogId=298955&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-298955
]
ASF GitHub Bot logged work on BEAM-7520:
----------------------------------------
Author: ASF GitHub Bot
Created on: 21/Aug/19 19:56
Start Date: 21/Aug/19 19:56
Worklog Time Spent: 10m
Work Description: je-ik commented on pull request #9190: [BEAM-7520] Fix
timer firing order in DirectRunner
URL: https://github.com/apache/beam/pull/9190#discussion_r316373884
##########
File path:
runners/direct-java/src/main/java/org/apache/beam/runners/direct/CommittedResult.java
##########
@@ -18,10 +18,10 @@
package org.apache.beam.runners.direct;
import com.google.auto.value.AutoValue;
+import java.util.Optional;
Review comment:
Yes, I also think that the performance impact of creating one short lived
instance should be negligible. I actually tried to measure it, but I didn't get
any statistically significant difference in this case. Although, I have seen
significant impact in other cases (getting rid of creating useless instance in
one place of spark runner gave me about 20% performance gain, but that could
have been very specific case). But that is probably not that significant to
this PR. :-)
To conclude this, I will drop the commit where I changed the Optional in
DirectRunner and start a discuss thread related to making the usage consistent
across the code base. Currently there is approx 3:2 usage of java.util vs.
guava.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
Issue Time Tracking
-------------------
Worklog Id: (was: 298955)
Time Spent: 8.5h (was: 8h 20m)
> DirectRunner timers are not strictly time ordered
> -------------------------------------------------
>
> Key: BEAM-7520
> URL: https://issues.apache.org/jira/browse/BEAM-7520
> Project: Beam
> Issue Type: Bug
> Components: runner-direct
> Affects Versions: 2.13.0
> Reporter: Jan Lukavský
> Assignee: Jan Lukavský
> Priority: Major
> Time Spent: 8.5h
> Remaining Estimate: 0h
>
> Let's suppose we have the following situation:
> - statful ParDo with two timers - timerA and timerB
> - timerA is set for window.maxTimestamp() + 1
> - timerB is set anywhere between <windowStart, windowEnd), let's denote that
> timerB.timestamp
> - input watermark moves to BoundedWindow.TIMESTAMP_MAX_VALUE
> Then the order of timers is as follows (correct):
> - timerB
> - timerA
> But, if timerB sets another timer (say for timerB.timestamp + 1), then the
> order of timers will be:
> - timerB (timerB.timestamp)
> - timerA (BoundedWindow.TIMESTAMP_MAX_VALUE)
> - timerB (timerB.timestamp + 1)
> Which is not ordered by timestamp. The reason for this is that when the input
> watermark update is evaluated, the WatermarkManager,extractFiredTimers() will
> produce both timerA and timerB. That would be correct, but when timerB sets
> another timer, that breaks this.
--
This message was sent by Atlassian Jira
(v8.3.2#803003)
