[jira] [Created] (FLINK-18729) kafka
shengjk1 created FLINK-18729: Summary: kafka Key: FLINK-18729 URL: https://issues.apache.org/jira/browse/FLINK-18729 Project: Flink Issue Type: Improvement Reporter: shengjk1 -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) kafka
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Component/s: Connectors / Kafka Affects Version/s: 1.10.0 1.10.1 1.11.0 1.11.1 Description: Now, flink kafka producer can't auto discovery partition when the change of kafka partition. maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis > kafka > - > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > Now, flink kafka producer can't auto discovery partition when the change of > kafka partition. maybe we can improvement it, use parameter: > flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (FLINK-18729) kafka
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17165593#comment-17165593 ] shengjk1 commented on FLINK-18729: -- if no any question, i will write code about it. > kafka > - > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > Now, flink kafka producer can't auto discovery partition when the change of > kafka partition. maybe we can improvement it, use parameter: > flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) kafka
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Description: Now, flink kafka producer can't auto discovery partition when the change of kafka partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis (was: Now, flink kafka producer can't auto discovery partition when the change of kafka partition. maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis) > kafka > - > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > Now, flink kafka producer can't auto discovery partition when the change of > kafka partition. Maybe we can improvement it, use parameter: > flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) Flink kafka producer auto discovery partition
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Summary: Flink kafka producer auto discovery partition (was: kafka) > Flink kafka producer auto discovery partition > - > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > Now, flink kafka producer can't auto discovery partition when the change of > kafka partition. Maybe we can improvement it, use parameter: > flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) Flink streaming kafka producer auto discovery partition
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Description: Now, flink streaming kafka producer can't auto discovery partition when the change of kafka partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis (was: Now, flink kafka producer can't auto discovery partition when the change of kafka partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis) Summary: Flink streaming kafka producer auto discovery partition (was: Flink kafka producer auto discovery partition) > Flink streaming kafka producer auto discovery partition > --- > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > Now, flink streaming kafka producer can't auto discovery partition when the > change of kafka partition. Maybe we can improvement it, use parameter: > flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) Flink streaming kafka producer auto discovery partition
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Description: Now, flink streaming kafka producer can't auto discovery partition when the change of kafka partition, such as add partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis (was: Now, flink streaming kafka producer can't auto discovery partition when the change of kafka partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis) > Flink streaming kafka producer auto discovery partition > --- > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > Now, flink streaming kafka producer can't auto discovery partition when the > change of kafka partition, such as add partition. Maybe we can improvement > it, use parameter: flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) Flink streaming kafka producer auto discovery partition
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Description: Now, In streaming api, flink kafka producer can't auto discovery partition when the change of kafka partition, such as add partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis (was: Now, flink streaming kafka producer can't auto discovery partition when the change of kafka partition, such as add partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis) > Flink streaming kafka producer auto discovery partition > --- > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > Now, In streaming api, flink kafka producer can't auto discovery partition > when the change of kafka partition, such as add partition. Maybe we can > improvement it, use parameter: flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) Make flink streaming kafka producer auto discovery partition
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Summary: Make flink streaming kafka producer auto discovery partition (was: Flink streaming kafka producer auto discovery partition) > Make flink streaming kafka producer auto discovery partition > > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > In streaming api, flink kafka producer can't auto discovery partition when > the change of kafka partition, such as add partition. Maybe we can > improvement it, use parameter: flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Updated] (FLINK-18729) Flink streaming kafka producer auto discovery partition
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-18729: - Description: In streaming api, flink kafka producer can't auto discovery partition when the change of kafka partition, such as add partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis (was: Now, In streaming api, flink kafka producer can't auto discovery partition when the change of kafka partition, such as add partition. Maybe we can improvement it, use parameter: flink.partition-discovery.interval-millis) > Flink streaming kafka producer auto discovery partition > --- > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > In streaming api, flink kafka producer can't auto discovery partition when > the change of kafka partition, such as add partition. Maybe we can > improvement it, use parameter: flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-18729) Make flink streaming kafka producer auto discovery partition
[ https://issues.apache.org/jira/browse/FLINK-18729?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17165593#comment-17165593 ] shengjk1 edited comment on FLINK-18729 at 7/27/20, 11:53 AM: - if no any question, i will write code and modify document. was (Author: shengjk1): if no any question, i will write code about it. > Make flink streaming kafka producer auto discovery partition > > > Key: FLINK-18729 > URL: https://issues.apache.org/jira/browse/FLINK-18729 > Project: Flink > Issue Type: Improvement > Components: Connectors / Kafka >Affects Versions: 1.10.0, 1.10.1, 1.11.0, 1.11.1 >Reporter: shengjk1 >Priority: Major > > In streaming api, flink kafka producer can't auto discovery partition when > the change of kafka partition, such as add partition. Maybe we can > improvement it, use parameter: flink.partition-discovery.interval-millis -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Created] (FLINK-11469) fix Tuning Checkpoints and Large State doc
shengjk1 created FLINK-11469: Summary: fix Tuning Checkpoints and Large State doc Key: FLINK-11469 URL: https://issues.apache.org/jira/browse/FLINK-11469 Project: Flink Issue Type: Improvement Components: Documentation Affects Versions: 1.7.1, 1.7.0, 1.6.3, 1.6.2, 1.6.4, 1.7.2, 1.8.0 Reporter: shengjk1 Sample code for subtitle Tuning RocksDB in Tuning Checkpoints and Large State is wrong Affects Version:All versions after 1.1 -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Assigned] (FLINK-11608) Translate the "Local Setup Tutorial" page into Chinese
[ https://issues.apache.org/jira/browse/FLINK-11608?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 reassigned FLINK-11608: Assignee: shengjk1 > Translate the "Local Setup Tutorial" page into Chinese > -- > > Key: FLINK-11608 > URL: https://issues.apache.org/jira/browse/FLINK-11608 > Project: Flink > Issue Type: Sub-task > Components: chinese-translation, Documentation >Reporter: Jark Wu >Assignee: shengjk1 >Priority: Major > > The page url is > https://ci.apache.org/projects/flink/flink-docs-master/tutorials/local_setup.html > The markdown file is located in flink/docs/tutorials/local_setup.zh.md > The markdown file will be created once FLINK-11529 is merged. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11608) Translate the "Local Setup Tutorial" page into Chinese
[ https://issues.apache.org/jira/browse/FLINK-11608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16770027#comment-16770027 ] shengjk1 commented on FLINK-11608: -- hi [~jark] I want to know when and where I can create a PR for it. where: fling-web repostitories or flink repositories or others when: PR can only be created after FLINK-11529 is merged or any time > Translate the "Local Setup Tutorial" page into Chinese > -- > > Key: FLINK-11608 > URL: https://issues.apache.org/jira/browse/FLINK-11608 > Project: Flink > Issue Type: Sub-task > Components: chinese-translation, Documentation >Reporter: Jark Wu >Assignee: shengjk1 >Priority: Major > > The page url is > https://ci.apache.org/projects/flink/flink-docs-master/tutorials/local_setup.html > The markdown file is located in flink/docs/tutorials/local_setup.zh.md > The markdown file will be created once FLINK-11529 is merged. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (FLINK-11608) Translate the "Local Setup Tutorial" page into Chinese
[ https://issues.apache.org/jira/browse/FLINK-11608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16770027#comment-16770027 ] shengjk1 edited comment on FLINK-11608 at 2/16/19 6:58 AM: --- hi [~jark] I want to know when and where I can create a PR for it. because i not find something about local_setup.md in flink-web and think create PR for this in flink repositories is not a wise move where: fling-web repostitories or flink repositories or others when: PR can only be created after FLINK-11529 is merged or any time was (Author: shengjk1): hi [~jark] I want to know when and where I can create a PR for it. where: fling-web repostitories or flink repositories or others when: PR can only be created after FLINK-11529 is merged or any time > Translate the "Local Setup Tutorial" page into Chinese > -- > > Key: FLINK-11608 > URL: https://issues.apache.org/jira/browse/FLINK-11608 > Project: Flink > Issue Type: Sub-task > Components: chinese-translation, Documentation >Reporter: Jark Wu >Assignee: shengjk1 >Priority: Major > > The page url is > https://ci.apache.org/projects/flink/flink-docs-master/tutorials/local_setup.html > The markdown file is located in flink/docs/tutorials/local_setup.zh.md > The markdown file will be created once FLINK-11529 is merged. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Created] (FLINK-11672) Add example for streaming operator connect
shengjk1 created FLINK-11672: Summary: Add example for streaming operator connect Key: FLINK-11672 URL: https://issues.apache.org/jira/browse/FLINK-11672 Project: Flink Issue Type: Improvement Reporter: shengjk1 Assignee: shengjk1 add example for streaming operator in code -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Created] (FLINK-11673) add example for streaming operators's broadcast
shengjk1 created FLINK-11673: Summary: add example for streaming operators's broadcast Key: FLINK-11673 URL: https://issues.apache.org/jira/browse/FLINK-11673 Project: Flink Issue Type: Improvement Components: Examples Reporter: shengjk1 Assignee: shengjk1 add example for streaming operators's broadcast in code -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11672) Add example for streaming operator connect
[ https://issues.apache.org/jira/browse/FLINK-11672?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11672: - Component/s: Examples > Add example for streaming operator connect > > > Key: FLINK-11672 > URL: https://issues.apache.org/jira/browse/FLINK-11672 > Project: Flink > Issue Type: Improvement > Components: Examples >Reporter: shengjk1 >Assignee: shengjk1 >Priority: Major > > add example for streaming operator in code -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11672) Add example for streaming operator connect
[ https://issues.apache.org/jira/browse/FLINK-11672?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11672: - Description: add example for streaming operator connect in code (was: add example for streaming operator in code) > Add example for streaming operator connect > > > Key: FLINK-11672 > URL: https://issues.apache.org/jira/browse/FLINK-11672 > Project: Flink > Issue Type: Improvement > Components: Examples >Reporter: shengjk1 >Assignee: shengjk1 >Priority: Major > > add example for streaming operator connect in code -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11672) Add example for streaming operator connect
[ https://issues.apache.org/jira/browse/FLINK-11672?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16773601#comment-16773601 ] shengjk1 commented on FLINK-11672: -- hi, [~StephanEwen] ,Thanks for your attention, I mean to add example for streaming operators's connect {{ DataStream1.connect({{DataStream2}}) }}, This is the error I expressed, i will update it. > Add example for streaming operator connect > > > Key: FLINK-11672 > URL: https://issues.apache.org/jira/browse/FLINK-11672 > Project: Flink > Issue Type: Improvement > Components: Examples >Reporter: shengjk1 >Assignee: shengjk1 >Priority: Major > > add example for streaming operator connect in code -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11672) Add example for streaming operators's connect
[ https://issues.apache.org/jira/browse/FLINK-11672?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11672: - Description: add example for streaming operators's connect such as \{{datastream1.connect(datastream2)}} in code (was: add example for streaming operator connect in code) Summary: Add example for streaming operators's connect(was: Add example for streaming operator connect ) > Add example for streaming operators's connect > --- > > Key: FLINK-11672 > URL: https://issues.apache.org/jira/browse/FLINK-11672 > Project: Flink > Issue Type: Improvement > Components: Examples >Reporter: shengjk1 >Assignee: shengjk1 >Priority: Major > > add example for streaming operators's connect such as > \{{datastream1.connect(datastream2)}} in code -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16774999#comment-16774999 ] shengjk1 commented on FLINK-11725: -- i think two ways solve this issue: 1.Add judgment in org.apache.flink.client.cli.CliFrontend if \{{-yn}} * \{{-ys}} < \{{-p}} system.exit(1) and Print message: \{{-yn}} * \{{-ys}} shoule more than the \{{-p}} 2.Create taskManager with one solt until \{{-yn}} * \{{-ys}} >= \{{-p}} as default > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.2 > linux > java8 > kafka1.1 >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Created] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
shengjk1 created FLINK-11725: Summary: use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic Key: FLINK-11725 URL: https://issues.apache.org/jira/browse/FLINK-11725 Project: Flink Issue Type: Bug Components: ResourceManager Affects Versions: 1.7.2 Environment: flink1.7.2 linux java8 kafka1.1 Reporter: shengjk1 Attachments: .png, 2.png, 3.png, MainConnect3.java I provide a simple example and used resource's picture for this issue this application is about flink consumer kafka, when submit job such as : flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & from yarn web ,i see, the total resources are erratic, and in most cases, all the remaining resources of the cluster will be used by the application -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16774999#comment-16774999 ] shengjk1 edited comment on FLINK-11725 at 2/22/19 10:53 AM: i think two ways solve this issue: 1).Add judgment in org.apache.flink.client.cli.CliFrontend if {{-yn}} * {{-ys}} < {{-p}} system.exit(1) and Print message: {{-yn}} * {{-ys}} shoule more than the {{-p}} 2).Create taskManager with one solt until {{-yn}} * {{-ys}} >= {{-p}} as default maybe 1) is more friendly to users was (Author: shengjk1): i think two ways solve this issue: 1.Add judgment in org.apache.flink.client.cli.CliFrontend if \{{-yn}} * \{{-ys}} < \{{-p}} system.exit(1) and Print message: \{{-yn}} * \{{-ys}} shoule more than the \{{-p}} 2.Create taskManager with one solt until \{{-yn}} * \{{-ys}} >= \{{-p}} as default > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.2 > linux > java8 > kafka1.1 >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11725: - Environment: flink1.7.1 linux java8 kafka1.1 was: flink1.7.2 linux java8 kafka1.1 > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.1 > linux > java8 > kafka1.1 >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11725: - Attachment: MainConnect4.java > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.1 > linux > java8 > kafka1.1 >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java, > MainConnect4.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16774999#comment-16774999 ] shengjk1 edited comment on FLINK-11725 at 2/25/19 2:54 AM: --- i think two ways solve this issue: 1).Add judgment about if {{-yn}} * {{-ys}} < {{-p}} system.exit(1) and Print message: {{-yn}} * {{-ys}} shoule more than the {{-p}} 2).Create taskManager with one solt until {{-yn}} * {{-ys}} >= {{-p}} as default maybe 1) is more friendly to users was (Author: shengjk1): i think two ways solve this issue: 1).Add judgment in org.apache.flink.client.cli.CliFrontend if {{-yn}} * {{-ys}} < {{-p}} system.exit(1) and Print message: {{-yn}} * {{-ys}} shoule more than the {{-p}} 2).Create taskManager with one solt until {{-yn}} * {{-ys}} >= {{-p}} as default maybe 1) is more friendly to users > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.1 > linux > java8 > kafka1.1 >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java, > MainConnect4.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11725: - Description: I provide a simple example and used resource's picture for this issue this application is about flink consumer kafka, when submit job such as : flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & or flink-1.7.1/bin/flink run -m yarn-cluster -ynm test -cMainConnect4 ./flinkDemo-1.0-SNAPSHOT.jar & from yarn web ,i see, the total resources are erratic, and in most cases, all the remaining resources of the cluster will be used by the application was: I provide a simple example and used resource's picture for this issue this application is about flink consumer kafka, when submit job such as : flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & from yarn web ,i see, the total resources are erratic, and in most cases, all the remaining resources of the cluster will be used by the application > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.1 > linux > java8 > kafka1.1 >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java, > MainConnect4.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > or > flink-1.7.1/bin/flink run -m yarn-cluster -ynm test -cMainConnect4 > ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11725: - Environment: flink1.7.2 linux java8 kafka1.1 scala 2.1 CDH was: flink1.7.1 linux java8 kafka1.1 > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.2 > linux > java8 > kafka1.1 > scala 2.1 > CDH >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java, > MainConnect4.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > or > flink-1.7.1/bin/flink run -m yarn-cluster -ynm test -cMainConnect4 > ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Closed] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 closed FLINK-11725. Resolution: Fixed Fix Version/s: 1.7.2 > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.2 > linux > java8 > kafka1.1 > scala 2.1 > CDH >Reporter: shengjk1 >Priority: Major > Fix For: 1.7.2 > > Attachments: .png, 2.png, 3.png, MainConnect3.java, > MainConnect4.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > or > flink-1.7.1/bin/flink run -m yarn-cluster -ynm test -cMainConnect4 > ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11725) use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are erratic
[ https://issues.apache.org/jira/browse/FLINK-11725?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16776504#comment-16776504 ] shengjk1 commented on FLINK-11725: -- 1.7.2已修复 > use flink cli when {{-yn}} * {{-ys}} < {{-p}}, the applied resources are > erratic > -- > > Key: FLINK-11725 > URL: https://issues.apache.org/jira/browse/FLINK-11725 > Project: Flink > Issue Type: Bug > Components: ResourceManager >Affects Versions: 1.7.2 > Environment: flink1.7.2 > linux > java8 > kafka1.1 > scala 2.1 > CDH >Reporter: shengjk1 >Priority: Major > Attachments: .png, 2.png, 3.png, MainConnect3.java, > MainConnect4.java > > > I provide a simple example and used resource's picture for this issue > this application is about flink consumer kafka, when submit job such as : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ynm test -p 5 > -cMainConnect3 ./flinkDemo-1.0-SNAPSHOT.jar & > or > flink-1.7.1/bin/flink run -m yarn-cluster -ynm test -cMainConnect4 > ./flinkDemo-1.0-SNAPSHOT.jar & > from yarn web ,i see, the total resources are erratic, and in most cases, all > the remaining resources of the cluster will be used by the application > -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11336) Flink HA didn't remove ZK metadata
[ https://issues.apache.org/jira/browse/FLINK-11336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16781618#comment-16781618 ] shengjk1 commented on FLINK-11336: -- hi, [~till.rohrmann] I have other questions and suggestions: 1. I want to know if will also delete invalid directories on HDFS, similar to zk metadata? because most of the metadata of HA is stored on HDFS. such as when job is failed. 2. when the job is canceled, the job's metadata is deleted as default , but i think it also should delete the corresponding directory, such as \{{jobId}}/shared and \{{jobId}}/taskowned. > Flink HA didn't remove ZK metadata > -- > > Key: FLINK-11336 > URL: https://issues.apache.org/jira/browse/FLINK-11336 > Project: Flink > Issue Type: Bug > Components: Runtime / Coordination >Reporter: shengjk1 >Assignee: Till Rohrmann >Priority: Major > Attachments: image-2019-01-15-19-42-21-902.png > > > Flink HA didn't remove ZK metadata > such as > go to zk cli : ls /flinkone > !image-2019-01-15-19-42-21-902.png! > > i suggest we should delete this metadata when the application cancel or > throw exception -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11336) Flink HA didn't remove ZK metadata
[ https://issues.apache.org/jira/browse/FLINK-11336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16781683#comment-16781683 ] shengjk1 commented on FLINK-11336: -- [~till.rohrmann] Yay, i think too, thank you > Flink HA didn't remove ZK metadata > -- > > Key: FLINK-11336 > URL: https://issues.apache.org/jira/browse/FLINK-11336 > Project: Flink > Issue Type: Bug > Components: Runtime / Coordination >Reporter: shengjk1 >Assignee: Till Rohrmann >Priority: Major > Attachments: image-2019-01-15-19-42-21-902.png > > > Flink HA didn't remove ZK metadata > such as > go to zk cli : ls /flinkone > !image-2019-01-15-19-42-21-902.png! > > i suggest we should delete this metadata when the application cancel or > throw exception -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11789) Checkpoint directories are not cleaned up after job termination
[ https://issues.apache.org/jira/browse/FLINK-11789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16781722#comment-16781722 ] shengjk1 commented on FLINK-11789: -- i think delete checkpoints/JOB_ID}} should contains three situations at least:}} 1. the job is killed by user 2. the job is canceled and the \{{ExternalizedCheckpointCleanup}} is \{{DELETE_ON_CANCELLATION}} 3. the job is fail If the user manually specifies \{{checkpoints/JOB_ID}} as savepoint dir , this directory should be kept, such as flink cancel -s checkpoints/JOB_ID > Checkpoint directories are not cleaned up after job termination > --- > > Key: FLINK-11789 > URL: https://issues.apache.org/jira/browse/FLINK-11789 > Project: Flink > Issue Type: Bug > Components: Runtime / Checkpointing >Affects Versions: 1.9.0 >Reporter: Till Rohrmann >Priority: Major > > Flink currently does not clean up all checkpoint directories when a job > reaches a globally terminal state. Having configured the checkpoint directory > {{checkpoints}}, I observe that after cancelling the job {{JOB_ID}} there are > still > {code} > checkpoints/JOB_ID/shared > checkpoints/JOB_ID/taskowned > {code} > I think it would be good if would delete {{checkpoints/JOB_ID}} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-7289: Affects Version/s: 1.7.2 1.8.2 1.9.0 > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory management in RocksDB can be found here: > https://github.com/facebook/rocksdb/wiki/Memory-usage-in-RocksDB > https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide > We should try to figure out ways to help users in one or more of the > following ways: > - Some way to autotune or calculate the RocksDB configuration. > - Conservative default values. > - Additional documentation. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (FLINK-11789) Checkpoint directories are not cleaned up after job termination
[ https://issues.apache.org/jira/browse/FLINK-11789?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16785178#comment-16785178 ] shengjk1 commented on FLINK-11789: -- i agree [~yunta] , I also think it would be worthy to discuss current checkpoint directory layout, suggest to create a separate thread , in order to not blow up the scope of this issue > Checkpoint directories are not cleaned up after job termination > --- > > Key: FLINK-11789 > URL: https://issues.apache.org/jira/browse/FLINK-11789 > Project: Flink > Issue Type: Bug > Components: Runtime / Checkpointing >Affects Versions: 1.9.0 >Reporter: Till Rohrmann >Priority: Major > > Flink currently does not clean up all checkpoint directories when a job > reaches a globally terminal state. Having configured the checkpoint directory > {{checkpoints}}, I observe that after cancelling the job {{JOB_ID}} there are > still > {code} > checkpoints/JOB_ID/shared > checkpoints/JOB_ID/taskowned > {code} > I think it would be good if would delete {{checkpoints/JOB_ID}} -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11608) Translate the "Local Setup Tutorial" page into Chinese
[ https://issues.apache.org/jira/browse/FLINK-11608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16905962#comment-16905962 ] shengjk1 commented on FLINK-11608: -- Hi [~WangHW] , thanks for your comments. I also read [“Flink Translation Specifications”|https://cwiki.apache.org/confluence/display/FLINK/Flink+Translation+Specifications], so we can start work. [~jark]. But I am also confused. I have read [submit your contribution|https://flink.apache.org/contributing/contribute-documentation.html#submit-your-contribution] , but i do not how to update *.md . Shoule i create a new PR or update and commit change in the old PR ? > Translate the "Local Setup Tutorial" page into Chinese > -- > > Key: FLINK-11608 > URL: https://issues.apache.org/jira/browse/FLINK-11608 > Project: Flink > Issue Type: Sub-task > Components: chinese-translation, Documentation >Reporter: Jark Wu >Assignee: shengjk1 >Priority: Major > Labels: pull-request-available > Time Spent: 10m > Remaining Estimate: 0h > > The page url is > https://ci.apache.org/projects/flink/flink-docs-master/tutorials/local_setup.html > The markdown file is located in flink/docs/tutorials/local_setup.zh.md > The markdown file will be created once FLINK-11530 is merged. -- This message was sent by Atlassian JIRA (v7.6.14#76016)
[jira] [Comment Edited] (FLINK-11608) Translate the "Local Setup Tutorial" page into Chinese
[ https://issues.apache.org/jira/browse/FLINK-11608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16905962#comment-16905962 ] shengjk1 edited comment on FLINK-11608 at 8/13/19 8:51 AM: --- Hi [~WangHW] , thanks for your comments. I also read [“Flink Translation Specifications”|https://cwiki.apache.org/confluence/display/FLINK/Flink+Translation+Specifications], so we can start work [~jark]. But I am also confused. I have read [submit your contribution|https://flink.apache.org/contributing/contribute-documentation.html#submit-your-contribution] , but i do not how to update *.md . Shoule i create a new PR or update and commit change in the old PR ? was (Author: shengjk1): Hi [~WangHW] , thanks for your comments. I also read [“Flink Translation Specifications”|https://cwiki.apache.org/confluence/display/FLINK/Flink+Translation+Specifications], so we can start work. [~jark]. But I am also confused. I have read [submit your contribution|https://flink.apache.org/contributing/contribute-documentation.html#submit-your-contribution] , but i do not how to update *.md . Shoule i create a new PR or update and commit change in the old PR ? > Translate the "Local Setup Tutorial" page into Chinese > -- > > Key: FLINK-11608 > URL: https://issues.apache.org/jira/browse/FLINK-11608 > Project: Flink > Issue Type: Sub-task > Components: chinese-translation, Documentation >Reporter: Jark Wu >Assignee: shengjk1 >Priority: Major > Labels: pull-request-available > Time Spent: 10m > Remaining Estimate: 0h > > The page url is > https://ci.apache.org/projects/flink/flink-docs-master/tutorials/local_setup.html > The markdown file is located in flink/docs/tutorials/local_setup.zh.md > The markdown file will be created once FLINK-11530 is merged. -- This message was sent by Atlassian JIRA (v7.6.14#76016)
[jira] [Commented] (FLINK-11608) Translate the "Local Setup Tutorial" page into Chinese
[ https://issues.apache.org/jira/browse/FLINK-11608?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16907048#comment-16907048 ] shengjk1 commented on FLINK-11608: -- Ok, I have appended an update commit [~jark] [~WangHW] > Translate the "Local Setup Tutorial" page into Chinese > -- > > Key: FLINK-11608 > URL: https://issues.apache.org/jira/browse/FLINK-11608 > Project: Flink > Issue Type: Sub-task > Components: chinese-translation, Documentation >Reporter: Jark Wu >Assignee: shengjk1 >Priority: Major > Labels: pull-request-available > Time Spent: 10m > Remaining Estimate: 0h > > The page url is > https://ci.apache.org/projects/flink/flink-docs-master/tutorials/local_setup.html > The markdown file is located in flink/docs/tutorials/local_setup.zh.md > The markdown file will be created once FLINK-11530 is merged. -- This message was sent by Atlassian JIRA (v7.6.14#76016)
[jira] [Created] (FLINK-11257) FlinkKafkaConsumer should support assgin partition
shengjk1 created FLINK-11257: Summary: FlinkKafkaConsumer should support assgin partition Key: FLINK-11257 URL: https://issues.apache.org/jira/browse/FLINK-11257 Project: Flink Issue Type: New Feature Components: Kafka Connector Affects Versions: 1.7.1 Reporter: shengjk1 i find flink 1.7 also has universal Kafka connector ,if the kakfa-connector support assgin partition ,the the kakfa-connector should prefect. such as a kafka topci has 3 partition, i only use 1 partition,but i should read all partition then filter.this method Not only waste resources but also relatively low efficiency.so i suggest FlinkKafkaConsumer should support assgin partition -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11257) FlinkKafkaConsumer should support assgin partition
[ https://issues.apache.org/jira/browse/FLINK-11257?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11257: - Priority: Major (was: Minor) > FlinkKafkaConsumer should support assgin partition > --- > > Key: FLINK-11257 > URL: https://issues.apache.org/jira/browse/FLINK-11257 > Project: Flink > Issue Type: New Feature > Components: Kafka Connector >Affects Versions: 1.7.1 >Reporter: shengjk1 >Priority: Major > > i find flink 1.7 also has universal Kafka connector ,if the kakfa-connector > support assgin partition ,the the kakfa-connector should prefect. such as a > kafka topci has 3 partition, i only use 1 partition,but i should read all > partition then filter.this method Not only waste resources but also > relatively low efficiency.so i suggest FlinkKafkaConsumer should support > assgin partition -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11257) FlinkKafkaConsumer should support assgin partition
[ https://issues.apache.org/jira/browse/FLINK-11257?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16732855#comment-16732855 ] shengjk1 commented on FLINK-11257: -- hi ,[~till.rohrmann],In order to ensure the compatibility of subsequent code. i think should change FlinkKafkaConsumerBase ,such as add one class KafkaTopicAndPartition and one interface KafkaDescriptor ,KafkaTopicAndPartition and KafkaTopicsDescriptor implements KafkaDescriptor,FlinkKafkaConsumerBase's topicsDescriptor change to KafkaDescriptor,then FlinkKafkaConsumerBase add one Constructor,this constructor modify this.kafkaDescriptor = new KafkaTopicsAndPartitionDescriptor(topicAndPartitions, topicAndPartitionPattern); may be have better ways ,welcome the discussion > FlinkKafkaConsumer should support assgin partition > --- > > Key: FLINK-11257 > URL: https://issues.apache.org/jira/browse/FLINK-11257 > Project: Flink > Issue Type: Sub-task > Components: Kafka Connector >Affects Versions: 1.7.1 >Reporter: shengjk1 >Assignee: vinoyang >Priority: Major > > i find flink 1.7 also has universal Kafka connector ,if the kakfa-connector > support assgin partition ,the the kakfa-connector should prefect. such as a > kafka topci has 3 partition, i only use 1 partition,but i should read all > partition then filter.this method Not only waste resources but also > relatively low efficiency.so i suggest FlinkKafkaConsumer should support > assgin partition -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Created] (FLINK-11336) Flink HA didn't remove ZK metadata
shengjk1 created FLINK-11336: Summary: Flink HA didn't remove ZK metadata Key: FLINK-11336 URL: https://issues.apache.org/jira/browse/FLINK-11336 Project: Flink Issue Type: Improvement Reporter: shengjk1 Attachments: image-2019-01-15-19-42-21-902.png Flink HA didn't remove ZK metadata such as go to zk cli : ls /flinkone !image-2019-01-15-19-42-21-902.png! i suggest we should delete this metadata when the application cancel or throw exception -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Created] (FLINK-11377) AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn if cli -yn -ys not effect
shengjk1 created FLINK-11377: Summary: AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn if cli -yn -ys not effect Key: FLINK-11377 URL: https://issues.apache.org/jira/browse/FLINK-11377 Project: Flink Issue Type: Bug Affects Versions: 1.7.1 Environment: java: jdk1.8.0_151 flink: flink-1.7.1 CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 Reporter: shengjk1 Attachments: image-2019-01-17-14-57-24-060.png when cli -yn -ys not effect ,AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn (cli -yn -ys not effect can refer to https://issues.apache.org/jira/browse/FLINK-11376) the cli : flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar AbstractYarnClusterDescriptor's log : org.apache.flink.yarn.AbstractYarnClusterDescriptor - Cluster specification: ClusterSpecification\{masterMemoryMB=1024, taskManagerMemoryMB=1024, numberTaskManagers=1, slotsPerTaskManager=1} but yarn web ui: !image-2019-01-17-14-57-24-060.png! -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Created] (FLINK-11376) flink cli -yn -ys is not effect if (yn * ys)
shengjk1 created FLINK-11376: Summary: flink cli -yn -ys is not effect if (yn * ys)https://issues.apache.org/jira/browse/FLINK-11376 Project: Flink Issue Type: Bug Affects Versions: 1.7.1 Environment: java: jdk1.8.0_151 flink: flink-1.7.1 CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 Reporter: shengjk1 Attachments: Main222.java, image-2019-01-17-14-25-34-206.png Such as the title if (yn * ys)parallelism ,the yn、ys will effect. if code write env.setParallelism(parallelism) ,the final application resources are yn multiples and ys multiples. such as parallelism=10,yn=1 ys=5,the final application resources:cpu cores=11 containers=3 Reproduce for the convenience of bugs,offer codes -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11377) AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn if cli -yn -ys not effect
[ https://issues.apache.org/jira/browse/FLINK-11377?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11377: - Description: when cli -yn -ys not effect ,AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn (cli -yn -ys not effect can refer to https://issues.apache.org/jira/browse/FLINK-11376) the cli : flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar AbstractYarnClusterDescriptor's log : org.apache.flink.yarn.AbstractYarnClusterDescriptor - Cluster specification: ClusterSpecification\{masterMemoryMB=1024, taskManagerMemoryMB=1024, numberTaskManagers=1, slotsPerTaskManager=1} but yarn web ui: allocated containers=4 and allocated cpu cores=4 !image-2019-01-17-14-57-24-060.png! was: when cli -yn -ys not effect ,AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn (cli -yn -ys not effect can refer to https://issues.apache.org/jira/browse/FLINK-11376) the cli : flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar AbstractYarnClusterDescriptor's log : org.apache.flink.yarn.AbstractYarnClusterDescriptor - Cluster specification: ClusterSpecification\{masterMemoryMB=1024, taskManagerMemoryMB=1024, numberTaskManagers=1, slotsPerTaskManager=1} but yarn web ui: !image-2019-01-17-14-57-24-060.png! > AbstractYarnClusterDescriptor's validClusterSpecification is not final > application resources from yarn if cli -yn -ys not effect > > > Key: FLINK-11377 > URL: https://issues.apache.org/jira/browse/FLINK-11377 > Project: Flink > Issue Type: Bug >Affects Versions: 1.7.1 > Environment: java: jdk1.8.0_151 > flink: flink-1.7.1 > CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-17-14-57-24-060.png > > > when cli -yn -ys not effect ,AbstractYarnClusterDescriptor's > validClusterSpecification is not final application resources from yarn (cli > -yn -ys not effect can refer to > https://issues.apache.org/jira/browse/FLINK-11376) > > the cli : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test > -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar > AbstractYarnClusterDescriptor's log : > org.apache.flink.yarn.AbstractYarnClusterDescriptor - Cluster > specification: ClusterSpecification\{masterMemoryMB=1024, > taskManagerMemoryMB=1024, numberTaskManagers=1, slotsPerTaskManager=1} > but yarn web ui: > allocated containers=4 and allocated cpu cores=4 > !image-2019-01-17-14-57-24-060.png! -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11376) flink cli -yn -ys is not effect if (yn * ys)
[ https://issues.apache.org/jira/browse/FLINK-11376?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11376: - Attachment: (was: Main222.java) > flink cli -yn -ys is not effect if (yn * ys) parallelism form env.setParallelism(parallelism) ) ; > - > > Key: FLINK-11376 > URL: https://issues.apache.org/jira/browse/FLINK-11376 > Project: Flink > Issue Type: Bug >Affects Versions: 1.7.1 > Environment: java: jdk1.8.0_151 > flink: flink-1.7.1 > CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-17-14-25-34-206.png > > > Such as the title > if (yn * ys) env.setParallelism(parallelism) ) the yn and ys is not effect > my application is flink streaming read kafka . this kafka topic has 3 > partition,and setParallelism(3) in code.when i use cli submiitjobs > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test > -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar > the application apply for 4 cpu cores and 4 containers from yarn web ui > !image-2019-01-17-14-25-34-206.png! > but if code is not write env.setParallelism(parallelism) or > (yn*ys)>parallelism ,the yn、ys will effect. if code write > env.setParallelism(parallelism) ,the final application resources are yn > multiples and ys multiples. such as parallelism=10,yn=1 ys=5,the final > application resources:cpu cores=11 containers=3 > > Reproduce for the convenience of bugs,offer codes -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Updated] (FLINK-11376) flink cli -yn -ys is not effect if (yn * ys)
[ https://issues.apache.org/jira/browse/FLINK-11376?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-11376: - Attachment: Main222.java > flink cli -yn -ys is not effect if (yn * ys) parallelism form env.setParallelism(parallelism) ) ; > - > > Key: FLINK-11376 > URL: https://issues.apache.org/jira/browse/FLINK-11376 > Project: Flink > Issue Type: Bug >Affects Versions: 1.7.1 > Environment: java: jdk1.8.0_151 > flink: flink-1.7.1 > CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 >Reporter: shengjk1 >Priority: Major > Attachments: Main222.java, image-2019-01-17-14-25-34-206.png > > > Such as the title > if (yn * ys) env.setParallelism(parallelism) ) the yn and ys is not effect > my application is flink streaming read kafka . this kafka topic has 3 > partition,and setParallelism(3) in code.when i use cli submiitjobs > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test > -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar > the application apply for 4 cpu cores and 4 containers from yarn web ui > !image-2019-01-17-14-25-34-206.png! > but if code is not write env.setParallelism(parallelism) or > (yn*ys)>parallelism ,the yn、ys will effect. if code write > env.setParallelism(parallelism) ,the final application resources are yn > multiples and ys multiples. such as parallelism=10,yn=1 ys=5,the final > application resources:cpu cores=11 containers=3 > > Reproduce for the convenience of bugs,offer codes -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11376) flink cli -yn -ys is not effect if (yn * ys)
[ https://issues.apache.org/jira/browse/FLINK-11376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16745721#comment-16745721 ] shengjk1 commented on FLINK-11376: -- Hi,Till Rohrmann,thanks for you comment,I agree with your suggestion and revisited the flink source code ,i think it is not a bug.I will close all about the issue thanks shengjk1 On 01/17/2019 21:41,Till Rohrmann (JIRA) wrote: [ https://issues.apache.org/jira/browse/FLINK-11376?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16745055#comment-16745055 ] Till Rohrmann commented on FLINK-11376: --- Hi [~shengjk1], I don't fully understand the problem description. One comment, though, {{-yn}} will only be respected by Flink's legacy mode. The legacy mode has been removed with Flink 1.7. I would recommend not setting the parallelism in code and in stead use the {{-p}} flag of Flink's CLI to specify the desired parallelism. flink cli -yn -ys is not effect if (yn * ys)https://issues.apache.org/jira/browse/FLINK-11376 Project: Flink Issue Type: Bug Affects Versions: 1.7.1 Environment: java: jdk1.8.0_151 flink: flink-1.7.1 CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 Reporter: shengjk1 Priority: Major Attachments: Main222.java, image-2019-01-17-14-25-34-206.png Such as the title if (yn * ys)parallelism ,the yn、ys will effect. if code write env.setParallelism(parallelism) ,the final application resources are yn multiples and ys multiples. such as parallelism=10,yn=1 ys=5,the final application resources:cpu cores=11 containers=3 Reproduce for the convenience of bugs,offer codes -- This message was sent by Atlassian JIRA (v7.6.3#76005) > flink cli -yn -ys is not effect if (yn * ys) parallelism form env.setParallelism(parallelism) ) ; > - > > Key: FLINK-11376 > URL: https://issues.apache.org/jira/browse/FLINK-11376 > Project: Flink > Issue Type: Bug >Affects Versions: 1.7.1 > Environment: java: jdk1.8.0_151 > flink: flink-1.7.1 > CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 >Reporter: shengjk1 >Priority: Major > Attachments: Main222.java, image-2019-01-17-14-25-34-206.png > > > Such as the title > if (yn * ys) env.setParallelism(parallelism) ) the yn and ys is not effect > my application is flink streaming read kafka . this kafka topic has 3 > partition,and setParallelism(3) in code.when i use cli submiitjobs > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test > -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar > the application apply for 4 cpu cores and 4 containers from yarn web ui > !image-2019-01-17-14-25-34-206.png! > but if code is not write env.setParallelism(parallelism) or > (yn*ys)>parallelism ,the yn、ys will effect. if code write > env.setParallelism(parallelism) ,the final application resources are yn > multiples and ys multiples. such as parallelism=10,yn=1 ys=5,the final > application resources:cpu cores=11 containers=3 > > Reproduce for the convenience of bugs,offer codes -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Closed] (FLINK-11377) AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn if cli -yn -ys not effect
[ https://issues.apache.org/jira/browse/FLINK-11377?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 closed FLINK-11377. Resolution: Not A Problem > AbstractYarnClusterDescriptor's validClusterSpecification is not final > application resources from yarn if cli -yn -ys not effect > > > Key: FLINK-11377 > URL: https://issues.apache.org/jira/browse/FLINK-11377 > Project: Flink > Issue Type: Bug >Affects Versions: 1.7.1 > Environment: java: jdk1.8.0_151 > flink: flink-1.7.1 > CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-17-14-57-24-060.png > > > when cli -yn -ys not effect ,AbstractYarnClusterDescriptor's > validClusterSpecification is not final application resources from yarn (cli > -yn -ys not effect can refer to > https://issues.apache.org/jira/browse/FLINK-11376) > > the cli : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test > -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar > AbstractYarnClusterDescriptor's log : > org.apache.flink.yarn.AbstractYarnClusterDescriptor - Cluster > specification: ClusterSpecification\{masterMemoryMB=1024, > taskManagerMemoryMB=1024, numberTaskManagers=1, slotsPerTaskManager=1} > but yarn web ui: > allocated containers=4 and allocated cpu cores=4 > !image-2019-01-17-14-57-24-060.png! -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11377) AbstractYarnClusterDescriptor's validClusterSpecification is not final application resources from yarn if cli -yn -ys not effect
[ https://issues.apache.org/jira/browse/FLINK-11377?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16745724#comment-16745724 ] shengjk1 commented on FLINK-11377: -- yes thanks > AbstractYarnClusterDescriptor's validClusterSpecification is not final > application resources from yarn if cli -yn -ys not effect > > > Key: FLINK-11377 > URL: https://issues.apache.org/jira/browse/FLINK-11377 > Project: Flink > Issue Type: Bug >Affects Versions: 1.7.1 > Environment: java: jdk1.8.0_151 > flink: flink-1.7.1 > CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-17-14-57-24-060.png > > > when cli -yn -ys not effect ,AbstractYarnClusterDescriptor's > validClusterSpecification is not final application resources from yarn (cli > -yn -ys not effect can refer to > https://issues.apache.org/jira/browse/FLINK-11376) > > the cli : > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test > -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar > AbstractYarnClusterDescriptor's log : > org.apache.flink.yarn.AbstractYarnClusterDescriptor - Cluster > specification: ClusterSpecification\{masterMemoryMB=1024, > taskManagerMemoryMB=1024, numberTaskManagers=1, slotsPerTaskManager=1} > but yarn web ui: > allocated containers=4 and allocated cpu cores=4 > !image-2019-01-17-14-57-24-060.png! -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Closed] (FLINK-11376) flink cli -yn -ys is not effect if (yn * ys)
[ https://issues.apache.org/jira/browse/FLINK-11376?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 closed FLINK-11376. Resolution: Not A Problem > flink cli -yn -ys is not effect if (yn * ys) parallelism form env.setParallelism(parallelism) ) ; > - > > Key: FLINK-11376 > URL: https://issues.apache.org/jira/browse/FLINK-11376 > Project: Flink > Issue Type: Bug >Affects Versions: 1.7.1 > Environment: java: jdk1.8.0_151 > flink: flink-1.7.1 > CDH:CDH-5.13.1-1.cdh5.13.1.p0.2 >Reporter: shengjk1 >Priority: Major > Attachments: Main222.java, image-2019-01-17-14-25-34-206.png > > > Such as the title > if (yn * ys) env.setParallelism(parallelism) ) the yn and ys is not effect > my application is flink streaming read kafka . this kafka topic has 3 > partition,and setParallelism(3) in code.when i use cli submiitjobs > flink-1.7.1/bin/flink run -m yarn-cluster -yn 1 -ys 1 -ynm test > -ccom.ishansong.bigdata.Main222 ./flinkDemo-1.0-SNAPSHOT.jar > the application apply for 4 cpu cores and 4 containers from yarn web ui > !image-2019-01-17-14-25-34-206.png! > but if code is not write env.setParallelism(parallelism) or > (yn*ys)>parallelism ,the yn、ys will effect. if code write > env.setParallelism(parallelism) ,the final application resources are yn > multiples and ys multiples. such as parallelism=10,yn=1 ys=5,the final > application resources:cpu cores=11 containers=3 > > Reproduce for the convenience of bugs,offer codes -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11336) Flink HA didn't remove ZK metadata
[ https://issues.apache.org/jira/browse/FLINK-11336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16745782#comment-16745782 ] shengjk1 commented on FLINK-11336: -- Unfamiliar with batch and bounded streams,so Inconvenient conclusion but such as unbounded streams when failed with no further retries cancelled we can remove the metadata ,As for how to start, you can start normally.I have already tried it, no problems in 1.8.0_151 flink 1.7.1 CDH5.13.1 > Flink HA didn't remove ZK metadata > -- > > Key: FLINK-11336 > URL: https://issues.apache.org/jira/browse/FLINK-11336 > Project: Flink > Issue Type: Improvement >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-15-19-42-21-902.png > > > Flink HA didn't remove ZK metadata > such as > go to zk cli : ls /flinkone > !image-2019-01-15-19-42-21-902.png! > > i suggest we should delete this metadata when the application cancel or > throw exception -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11336) Flink HA didn't remove ZK metadata
[ https://issues.apache.org/jira/browse/FLINK-11336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16747322#comment-16747322 ] shengjk1 commented on FLINK-11336: -- 1.No matter what form stop flink, such as cancel,failed with no further retries,kill, metadata not be deleted. 2.when cancel,failed with no further retries,kill,manually deleting metadata has no effect on newly launched programs even if there has a savepoint this is my observed behavior > Flink HA didn't remove ZK metadata > -- > > Key: FLINK-11336 > URL: https://issues.apache.org/jira/browse/FLINK-11336 > Project: Flink > Issue Type: Improvement >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-15-19-42-21-902.png > > > Flink HA didn't remove ZK metadata > such as > go to zk cli : ls /flinkone > !image-2019-01-15-19-42-21-902.png! > > i suggest we should delete this metadata when the application cancel or > throw exception -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Comment Edited] (FLINK-11336) Flink HA didn't remove ZK metadata
[ https://issues.apache.org/jira/browse/FLINK-11336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16749548#comment-16749548 ] shengjk1 edited comment on FLINK-11336 at 1/23/19 5:53 AM: --- Yarn (per job or as a session) was (Author: shengjk1): Yarn (per job or as a session) > Flink HA didn't remove ZK metadata > -- > > Key: FLINK-11336 > URL: https://issues.apache.org/jira/browse/FLINK-11336 > Project: Flink > Issue Type: Improvement >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-15-19-42-21-902.png > > > Flink HA didn't remove ZK metadata > such as > go to zk cli : ls /flinkone > !image-2019-01-15-19-42-21-902.png! > > i suggest we should delete this metadata when the application cancel or > throw exception -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-11336) Flink HA didn't remove ZK metadata
[ https://issues.apache.org/jira/browse/FLINK-11336?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16749548#comment-16749548 ] shengjk1 commented on FLINK-11336: -- Yarn (per job or as a session) > Flink HA didn't remove ZK metadata > -- > > Key: FLINK-11336 > URL: https://issues.apache.org/jira/browse/FLINK-11336 > Project: Flink > Issue Type: Improvement >Reporter: shengjk1 >Priority: Major > Attachments: image-2019-01-15-19-42-21-902.png > > > Flink HA didn't remove ZK metadata > such as > go to zk cli : ls /flinkone > !image-2019-01-15-19-42-21-902.png! > > i suggest we should delete this metadata when the application cancel or > throw exception -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956749#comment-16956749 ] shengjk1 commented on FLINK-7289: - I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. class BackendOptions implements ConfigurableOptionsFactory \{ // private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) \{ return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024) .setBlockSize(4 * 32 * 1024)) ; } @Override public OptionsFactory configure(Configuration configuration) \{ return this; } } > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory management in RocksDB can be found here: > https://github.com/facebook/rocksdb/wiki/Memory-usage-in-RocksDB > https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide > We should try to figure out ways to help users in one or more of the > following ways: > - Some way to autotune or calculate the RocksDB configuration. > - Conservative default values. > - Additional documentation. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956749#comment-16956749 ] shengjk1 edited comment on FLINK-7289 at 10/22/19 7:09 AM: --- I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. {code:java} class BackendOptions implements ConfigurableOptionsFactory { private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) { return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024)n .setBlockSize(4 * 32 * 1024)) ; } @Override public OptionsFactory configure(Configuration configuration) { return this; } } {code} was (Author: shengjk1): I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. {code:java} class BackendOptions implements ConfigurableOptionsFactory { // private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) { return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024) // increases read amplification but decreases memory useage and space amplification .setBlockSize(4 * 32 * 1024)) ; } @Override public OptionsFactory configure(Configuration configuration) { return this; } } {code} > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run in
[jira] [Comment Edited] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956749#comment-16956749 ] shengjk1 edited comment on FLINK-7289 at 10/22/19 7:09 AM: --- I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. {code:java} class BackendOptions implements ConfigurableOptionsFactory { // private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) { return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024) // increases read amplification but decreases memory useage and space amplification .setBlockSize(4 * 32 * 1024)) ; } @Override public OptionsFactory configure(Configuration configuration) { return this; } } {code} was (Author: shengjk1): I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. class BackendOptions implements ConfigurableOptionsFactory \{ // private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) \{ return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024) .setBlockSize(4 * 32 * 1024)) ; } @Override public OptionsFactory configure(Configuration configuration) \{ return this; } } > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same p
[jira] [Comment Edited] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956749#comment-16956749 ] shengjk1 edited comment on FLINK-7289 at 10/22/19 7:10 AM: --- I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. {code:java} class BackendOptions implements ConfigurableOptionsFactory { private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) { return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024)n .setBlockSize(4 * 32 * 1024)); } @Override public OptionsFactory configure(Configuration configuration) { return this; } } {code} was (Author: shengjk1): I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. {code:java} class BackendOptions implements ConfigurableOptionsFactory { private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) { return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024)n .setBlockSize(4 * 32 * 1024)) ; } @Override public OptionsFactory configure(Configuration configuration) { return this; } } {code} > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory
[jira] [Comment Edited] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16956749#comment-16956749 ] shengjk1 edited comment on FLINK-7289 at 10/22/19 7:11 AM: --- I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. was (Author: shengjk1): I also rebuild flink and use writeBufferManager. For one specific job I'm working on, I went from a OOM kill every 5-7 hours to at least 20 hours of no OOM events . But finally it still OOM . So I think maybe we also need other parameters which easy to use to solve it. {code:java} class BackendOptions implements ConfigurableOptionsFactory { private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions .setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) { return currentOptions .setLevelCompactionDynamicLevelBytes(true) .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024)n .setBlockSize(4 * 32 * 1024)); } @Override public OptionsFactory configure(Configuration configuration) { return this; } } {code} > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory management in RocksDB can be found here: > https://github.com/facebook/rocksdb/wiki/Memory-usage-in-RocksDB > https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide > We should try to figure out ways to help users in one or more of the > following ways: > - Some way to autotune or calculate the RocksDB configuration. > - Conservative default values. > - Additional documentation. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16958725#comment-16958725 ] shengjk1 commented on FLINK-7289: - [~yunta] I don't know how to get the LOG file rockdb, this is my code : {code:java} class BackendOptions implements ConfigurableOptionsFactory { protected final static org.slf4j.Logger logger = LoggerFactory.getLogger(Main.class); // private static final WriteBufferManager writeBufferManager = new WriteBufferManager(1 << 30, new LRUCache(1 << 18)); @Override public DBOptions createDBOptions(DBOptions currentOptions) { return currentOptions.setInfoLogLevel(InfoLogLevel.INFO_LEVEL).setLogger(new Logger(currentOptions) { @Override protected void log(InfoLogLevel infoLogLevel, String s) { System.out.println("rockdb=="+s); logger.debug("rockdb =={}",s); } }).setMaxBackgroundJobs(4) .setUseFsync(false) .setMaxBackgroundFlushes(3) .setWriteBufferManager(writeBufferManager); } @Override public ColumnFamilyOptions createColumnOptions(ColumnFamilyOptions currentOptions) { return currentOptions .setMinWriteBufferNumberToMerge(2) .setMaxWriteBufferNumber(5) .setOptimizeFiltersForHits(true) .setMaxWriteBufferNumberToMaintain(3) .setTableFormatConfig( new BlockBasedTableConfig() .setCacheIndexAndFilterBlocks(true) .setCacheIndexAndFilterBlocksWithHighPriority(true) .setBlockCacheSize(256 * 1024 * 1024) // increases read amplification but decreases memory useage and space amplification .setBlockSize(4 * 32 * 1024)) ; } @Override public OptionsFactory configure(Configuration configuration) { return this; } } {code} This is my cli: {code:java} -p 3 -ytm 2048M -ytm early -m yarn-cluster -yD "state.backend.rocksdb.ttl.compaction.filter.enabled=true" {code} > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory management in RocksDB can be found here: > https://github.com/facebook/rocksdb/wiki/Memory-usage-in-RocksDB > https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide > We should try to figure out ways to help users in one or more of the > following ways: > - Some way to autotune or calculate the RocksDB configuration. > - Conservative default values. > - Additional documentation. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16961652#comment-16961652 ] shengjk1 edited comment on FLINK-7289 at 10/29/19 3:55 AM: --- Thanks [~yunta] and [~mikekap]. I have solved it. As Mike Kaplinskiy said, i put the jar which only include BackendOptions.class in the flink boot classpath - not in my application code classpath. was (Author: shengjk1): Thanks [~yunta] and [~mikekap]. I have solved it. As Mike Kaplinskiy said, i put the jar which only include BackendOptions.class in the flink boot classpath - not in my application code classpath . > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory management in RocksDB can be found here: > https://github.com/facebook/rocksdb/wiki/Memory-usage-in-RocksDB > https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide > We should try to figure out ways to help users in one or more of the > following ways: > - Some way to autotune or calculate the RocksDB configuration. > - Conservative default values. > - Additional documentation. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16961652#comment-16961652 ] shengjk1 commented on FLINK-7289: - Thanks [~yunta] and [~mikekap]. I have solved it. As Mike Kaplinskiy said, i put the jar which only include BackendOptions.class in the flink boot classpath - not in my application code classpath . > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory management in RocksDB can be found here: > https://github.com/facebook/rocksdb/wiki/Memory-usage-in-RocksDB > https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide > We should try to figure out ways to help users in one or more of the > following ways: > - Some way to autotune or calculate the RocksDB configuration. > - Conservative default values. > - Additional documentation. -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Issue Comment Deleted] (FLINK-7289) Memory allocation of RocksDB can be problematic in container environments
[ https://issues.apache.org/jira/browse/FLINK-7289?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ] shengjk1 updated FLINK-7289: Comment: was deleted (was: Thanks [~yunta] and [~mikekap]. I have solved it. As Mike Kaplinskiy said, i put the jar which only include BackendOptions.class in the flink boot classpath - not in my application code classpath.) > Memory allocation of RocksDB can be problematic in container environments > - > > Key: FLINK-7289 > URL: https://issues.apache.org/jira/browse/FLINK-7289 > Project: Flink > Issue Type: Improvement > Components: Runtime / State Backends >Affects Versions: 1.2.0, 1.3.0, 1.4.0, 1.7.2, 1.8.2, 1.9.0 >Reporter: Stefan Richter >Priority: Major > Fix For: 1.10.0 > > Attachments: completeRocksdbConfig.txt > > > Flink's RocksDB based state backend allocates native memory. The amount of > allocated memory by RocksDB is not under the control of Flink or the JVM and > can (theoretically) grow without limits. > In container environments, this can be problematic because the process can > exceed the memory budget of the container, and the process will get killed. > Currently, there is no other option than trusting RocksDB to be well behaved > and to follow its memory configurations. However, limiting RocksDB's memory > usage is not as easy as setting a single limit parameter. The memory limit is > determined by an interplay of several configuration parameters, which is > almost impossible to get right for users. Even worse, multiple RocksDB > instances can run inside the same process and make reasoning about the > configuration also dependent on the Flink job. > Some information about the memory management in RocksDB can be found here: > https://github.com/facebook/rocksdb/wiki/Memory-usage-in-RocksDB > https://github.com/facebook/rocksdb/wiki/RocksDB-Tuning-Guide > We should try to figure out ways to help users in one or more of the > following ways: > - Some way to autotune or calculate the RocksDB configuration. > - Conservative default values. > - Additional documentation. -- This message was sent by Atlassian Jira (v8.3.4#803005)