[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17097665#comment-17097665
]
Luke Cwik commented on BEAM-8944:
-
I was able to come up with a compromise that I believe should work well
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17097302#comment-17097302
]
Maximilian Michels commented on BEAM-8944:
--
[~lcwik] For the time being, do you think we could
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17074641#comment-17074641
]
Maximilian Michels commented on BEAM-8944:
--
Thank you for your thoughts Luke. I agree that the
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17073871#comment-17073871
]
Luke Cwik commented on BEAM-8944:
-
I was having difficulty in producing a solution that both created
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17073657#comment-17073657
]
Maximilian Michels commented on BEAM-8944:
--
My first though was to change the implementation to
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17073142#comment-17073142
]
Luke Cwik commented on BEAM-8944:
-
I took a look at the implementation of the UnboundedThreadPoolExecutor
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17072860#comment-17072860
]
Maximilian Michels commented on BEAM-8944:
--
I see that the usage pattern in the description is a
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17072842#comment-17072842
]
Luke Cwik commented on BEAM-8944:
-
The Python 3.5 implementation of the ThreadPoolExecutor:
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17072840#comment-17072840
]
Luke Cwik commented on BEAM-8944:
-
So the original analysis done in the description of the bug isn't
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17072693#comment-17072693
]
Maximilian Michels commented on BEAM-8944:
--
!checkpoint-duration.png!
yellow: 2.18.0 as
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17072103#comment-17072103
]
Maximilian Michels commented on BEAM-8944:
--
Essentially, yes. Especially, this is a concern for
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17071998#comment-17071998
]
Yichi Zhang commented on BEAM-8944:
---
[~mxm] the checkpoint duration is lower means that without
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17001802#comment-17001802
]
Yichi Zhang commented on BEAM-8944:
---
I think so
> Python SDK harness performance degradation with
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17001589#comment-17001589
]
Ahmet Altay commented on BEAM-8944:
---
Could this be closed after the cherry pick PR
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17000448#comment-17000448
]
Yichi Zhang commented on BEAM-8944:
---
then yeah, it'll affect python streaming jobs (which is only on
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17000445#comment-17000445
]
Udi Meiri commented on BEAM-8944:
-
The point of my question is to figure whether this issue is a 2.18
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17000443#comment-17000443
]
Udi Meiri commented on BEAM-8944:
-
By current I don't mean the current Beam release but the current
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17000313#comment-17000313
]
Udi Meiri commented on BEAM-8944:
-
Does this bug affect current production runners (such as Dataflow
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16994104#comment-16994104
]
Yichi Zhang commented on BEAM-8944:
---
Seems like the original change was to solve deadlock and stuckness
[
https://issues.apache.org/jira/browse/BEAM-8944?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16993141#comment-16993141
]
Yichi Zhang commented on BEAM-8944:
---
CC: [~angoenka]
> Python SDK harness performance degradation with
20 matches
Mail list logo
