[jira] [Commented] (FLINK-17842) Performance regression on 19.05.2020

2020-05-25 Thread Piotr Nowojski (Jira)


[ 
https://issues.apache.org/jira/browse/FLINK-17842?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17116033#comment-17116033
 ] 

Piotr Nowojski commented on FLINK-17842:


Performance regression was only partially solved:
http://codespeed.dak8s.net:8000/timeline/?ben=serializerHeavyString=2 
seems fine, but those are still showing some slow down:
http://codespeed.dak8s.net:8000/timeline/?ben=networkBroadcastThroughput=2
http://codespeed.dak8s.net:8000/timeline/?ben=networkThroughput.1000,100ms=2

> Performance regression on 19.05.2020
> 
>
> Key: FLINK-17842
> URL: https://issues.apache.org/jira/browse/FLINK-17842
> Project: Flink
>  Issue Type: Bug
>  Components: Benchmarks
>Affects Versions: 1.11.0
>Reporter: Piotr Nowojski
>Assignee: Piotr Nowojski
>Priority: Blocker
>  Labels: pull-request-available
> Fix For: 1.11.0
>
>
> There is a noticeable performance regression in many benchmarks:
> http://codespeed.dak8s.net:8000/timeline/?ben=serializerHeavyString=2
> http://codespeed.dak8s.net:8000/timeline/?ben=networkThroughput.1000,1ms=2
> http://codespeed.dak8s.net:8000/timeline/?ben=networkThroughput.100,100ms=2
> http://codespeed.dak8s.net:8000/timeline/?ben=globalWindow=2
> that happened on May 19th, probably between 260ef2c and 2f18138



--
This message was sent by Atlassian Jira
(v8.3.4#803005)


[jira] [Commented] (FLINK-17842) Performance regression on 19.05.2020

2020-05-20 Thread Piotr Nowojski (Jira)


[ 
https://issues.apache.org/jira/browse/FLINK-17842?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17112056#comment-17112056
 ] 

Piotr Nowojski commented on FLINK-17842:


Those are the commits in suspected range:

{noformat}
2f18138df2 [26 hours ago] [FLINK-17809][dist] Quote classpath and 
FLINK_CONF_DIR [Chesnay Schepler]
31ec497b96 [2 days ago] [FLINK-17763][dist] Properly handle log properties and 
spaces in scala-shell.sh [Chesnay Schepler]
2aacb62c29 [13 days ago] [FLINK-17547][task] Implement getUnconsumedSegment for 
spilled buffers [Roman Khachatryan]
54155744bd [13 days ago] [FLINK-17547][task] Use RefCountedFile in 
SpanningWrapper (todo: merge with next?) [Roman Khachatryan]
2fcc1fca7c [13 days ago] [FLINK-17547][task][hotfix] Move RefCountedFile to 
flink-core to use it in SpanningWrapper [Roman Khachatryan]
179de29f09 [13 days ago] [FLINK-17547][task][hotfix] Extract 
RefCountedFileWithStream from RefCountedFile Motivation: use RefCountedFile for 
reading as well. [Roman Khachatryan]
37f441a2fc [2 days ago] [FLINK-17547][task] Use iterator for unconsumed 
buffers. Motivation: support spilled records Changes: 1. change 
SpillingAdaptiveSpanningRecordDeserializer.getUnconsumedBuffer signature 2. 
adapt channel state persistence to new types [Roman Khachatryan]
824100e146 [8 days ago] [FLINK-17547][task][hotfix] Extract methods from 
RecordsDeserializer [Roman Khachatryan]
67d3eae6f1 [8 days ago] [FLINK-17547][task][hotfix] Fix compiler warnings in 
NonSpanningWrapper [Roman Khachatryan]
d7b29f7bb5 [2 weeks ago] [FLINK-17547][task][hotfix] Extract SpanningWrapper 
from SpillingAdaptiveSpanningRecordDeserializer (static inner class). As it is, 
no logical changes. [Roman Khachatryan]
6e3c5abf7b [2 weeks ago] [FLINK-17547][task][hotfix] Extract NonSpanningWrapper 
from SpillingAdaptiveSpanningRecordDeserializer (static inner class) As it is, 
no logical changes. [Roman Khachatryan]
8548d37df6 [13 days ago] [FLINK-17547][task][hotfix] Improve error handling 1 
catch one more invalid input in DataOutputSerializer.write 2 more informative 
error messages [Roman Khachatryan]
{noformat}

It means the regression was probably caused by FLINK-17547 CC 
[~roman_khachatryan]

> Performance regression on 19.05.2020
> 
>
> Key: FLINK-17842
> URL: https://issues.apache.org/jira/browse/FLINK-17842
> Project: Flink
>  Issue Type: Bug
>  Components: Benchmarks
>Affects Versions: 1.11.0
>Reporter: Piotr Nowojski
>Assignee: Piotr Nowojski
>Priority: Blocker
> Fix For: 1.11.0
>
>
> There is a noticeable performance regression in many benchmarks:
> http://codespeed.dak8s.net:8000/timeline/?ben=serializerHeavyString=2
> http://codespeed.dak8s.net:8000/timeline/?ben=networkThroughput.1000,1ms=2
> http://codespeed.dak8s.net:8000/timeline/?ben=networkThroughput.100,100ms=2
> http://codespeed.dak8s.net:8000/timeline/?ben=globalWindow=2
> that happened on May 19th, probably between 260ef2c and 2f18138



--
This message was sent by Atlassian Jira
(v8.3.4#803005)