[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018790#comment-17018790 ] Victor Wong edited comment on FLINK-15447 at 1/19/20 4:24 AM: -- [~rongr] for me #2 is most close to the main concern, but we do not want to share with others for fine-grain control on disk resource, but for a shared location is very prone to be disk full. There are some good points in HADOOP-2735: _Can we add -Djava.io.tmpdir="./tmp" somewhere ?_ _so that,_ _1) Tasks can utilize all disks when using tmp_ _2) Any undeleted tmp files will be deleted by the tasktracker when task(job?) is done._ was (Author: victor-wong): [~rongr] for me #2 is most close to the main concern, but we do not want to share with others not for fine-grain control on disk resource, but for a shared location is very prone to be disk full. There are some good points in HADOOP-2735: _Can we add -Djava.io.tmpdir="./tmp" somewhere ?_ _so that,_ _1) Tasks can utilize all disks when using tmp_ _2) Any undeleted tmp files will be deleted by the tasktracker when task(job?) is done._ > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018790#comment-17018790 ] Victor Wong edited comment on FLINK-15447 at 1/19/20 4:23 AM: -- [~rongr] for me #2 is most close to the main concern, but we do not want to share with others not for fine-grain control on disk resource, but for a shared location is very prone to be disk full. There are some good points in HADOOP-2735: _Can we add -Djava.io.tmpdir="./tmp" somewhere ?_ _so that,_ _1) Tasks can utilize all disks when using tmp_ _2) Any undeleted tmp files will be deleted by the tasktracker when task(job?) is done._ was (Author: victor-wong): [~rongr] for me #2 is most close to the main concern, but we do not want to share with others not for fine-grain control on disk resource, but for a shared location is very prone to be disk full. There are some good points in [HADOOP-2735|https://issues.apache.org/jira/browse/HADOOP-2735]: _Can we add -Djava.io.tmpdir="./tmp" somewhere ? so that, 1) Tasks can utilize all disks when using tmp 2) Any undeleted tmp files will be deleted by the tasktracker when task(job?) is done._ > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018645#comment-17018645 ] Rong Rong edited comment on FLINK-15447 at 1/18/20 5:02 PM: Ahh. I see your intention. Let me clarify what I understood. please correct me if I were wrong: * Both Flink-YARN (JM/TM) and some other 3rd party uses {{java.io.tmpdir}} which is a JVM system env key. * This means, potentially, whatever directory configured, all Java process will be sharing this directory as tmp folder. So could you clarify which is the main concern: 1. We don't want to pollute {{/tmp}} - which potentially will be used also by NON-JVM processes. 2. We want Flink JM/TM to NOT share with others JVM process or other YARN containers. If the above analysis is correct, For #1, we actually creates dedicate partitions to put {{/tmp}} in our YARN node, which resolves the issue. Not sure if this can be a solution on your case. For #2, yes I think the question is not easy to answer especially we want fine-grain control on disk resource. was (Author: rongr): Ahh. I see your intention. Let me clarify what I understood. please correct me if I were wrong: * Both Flink-YARN (JM/TM) and some other 3rd party uses {{java.io.tmpdir}} which is a JVM system env key. * This means, potentially, whatever directory configured, all Java process will be sharing this directory as tmp folder. So could you clarify which is the main concern: 1. The default key is set to {{/tmp}} - which potentially will be used also by NON-JVM process. 2. In addition, we also want Flink JM/TM to NOT share with others JVM process or other YARN containers. If the above analysis is correct, For #1, we actually creates dedicate partitions to put {{/tmp}} in our YARN node, which resolves the issue. Not sure if this can be a solution on your case. For #2, yes I think the question is not easy to answer especially we want fine-grain control on disk resource. > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018645#comment-17018645 ] Rong Rong edited comment on FLINK-15447 at 1/18/20 5:01 PM: Ahh. I see your intention. Let me clarify what I understood. please correct me if I were wrong: * Both Flink-YARN (JM/TM) and some other 3rd party uses {{java.io.tmpdir}} which is a JVM system env key. * This means, potentially, whatever directory configured, all Java process will be sharing this directory as tmp folder. So could you clarify which is the main concern: 1. The default key is set to {{/tmp}} - which potentially will be used also by NON-JVM process. 2. In addition, we also want Flink JM/TM to NOT share with others JVM process or other YARN containers. If the above analysis is correct, For #1, we actually creates dedicate partitions to put {{/tmp}} in our YARN node, which resolves the issue. Not sure if this can be a solution on your case. For #2, yes I think the question is not easy to answer especially we want fine-grain control on disk resource. was (Author: rongr): Ahh. I see your intention. Let me clarify what I understood. please correct me if I were wrong: * Both Flink-YARN (JM/TM) and some other 3rd party uses {{java.io.tmpdir}} which is a JVM system env key. * This means, potentially, whatever directory configured via system env, all Java process will be sharing this location. So could you clarify which is the main concern: 1. The default key is set to {{/tmp}} - which potentially will be used also by NON-JVM process. 2. In addition, we also want Flink JM/TM to NOT share with others JVM process or other YARN containers. If the above analysis is correct, For #1, we actually creates dedicate partitions to put {{/tmp}} in our YARN node, which resolves the issue. Not sure if this can be a solution on your case. For #2, yes I think the question is not easy to answer especially we want fine-grain control on disk resource. > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018645#comment-17018645 ] Rong Rong edited comment on FLINK-15447 at 1/18/20 5:01 PM: Ahh. I see your intention. Let me clarify what I understood. please correct me if I were wrong: * Both Flink-YARN (JM/TM) and some other 3rd party uses {{java.io.tmpdir}} which is a JVM system env key. * This means, potentially, whatever directory configured via system env, all Java process will be sharing this location. So could you clarify which is the main concern: 1. The default key is set to {{/tmp}} - which potentially will be used also by NON-JVM process. 2. In addition, we also want Flink JM/TM to NOT share with others JVM process or other YARN containers. If the above analysis is correct, For #1, we actually creates dedicate partitions to put {{/tmp}} in our YARN node, which resolves the issue. Not sure if this can be a solution on your case. For #2, yes I think the question is not easy to answer especially we want fine-grain control on disk resource. was (Author: rongr): Ahh. I see your intention. Let me clarify what I understood. please correct me if I were wrong: * Both Flink-YARN (JM/TM) and some other 3rd party uses {{java.io.tmpdir}} which is a JVM system env key. * This means potentially means, whatever directory configured via system env, all Java process will be sharing this location. So could you clarify which is the main concern: 1. The default key is set to {{/tmp}} - which potentially will be used also by NON-JVM process. 2. In addition, we also want Flink JM/TM to NOT share with others JVM process or other YARN containers. If the above analysis is correct, For #1, we actually creates dedicate partitions to put {{/tmp}} in our YARN node, which resolves the issue. Not sure if this can be a solution on your case. For #2, yes I think the question is not easy to answer especially we want fine-grain control on disk resource. > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018462#comment-17018462 ] Victor Wong edited comment on FLINK-15447 at 1/18/20 1:41 AM: -- [~rongr] Thanks for your reply! _Could you elaborate what does it mean by Flink-YARN default the value to /tmp? I am guessing you mean JVM default the value to /tmp ?_ --- Yes, I mean that Flink-YARN is using the default value of JVM, which is "/tmp". _So far the only place I can see in flink yarn code utilizing this key is .._ --- The third-party dependencies might utilize this key as well, as [~lzljs3620320] mentioned. My intention is similar to how [~xymaqingxiang] mentioned that setting the value to a tmp directory under the working directory of Yarn container. was (Author: victor-wong): [~rongr] Thanks for your reply! _Could you elaborate what does it mean by Flink-YARN default the value to /tmp? I am guessing you mean JVM default the value to /tmp ?_ --- Yes, I mean that Flink-YARN is using the default value of JVM, which is "/tmp". _So far the only place I can see in flink yarn code utilizing this key is: https://github.com/apache/flink/blob/release-1.10/flink-yarn/src/main/java/org/apache/flink/yarn/cli/FlinkYarnSessionCli.java#L899_ --- The third-party dependencies might utilize this key as well, as [~lzljs3620320] mentioned. My intention is similar to how [~xymaqingxiang] mentioned that setting the value to a tmp directory under the working directory of Yarn container. > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018462#comment-17018462 ] Victor Wong edited comment on FLINK-15447 at 1/18/20 1:40 AM: -- [~rongr] Thanks for your reply! _Could you elaborate what does it mean by Flink-YARN default the value to /tmp? I am guessing you mean JVM default the value to /tmp ?_ --- Yes, I mean that Flink-YARN is using the default value of JVM, which is "/tmp". _So far the only place I can see in flink yarn code utilizing this key is: https://github.com/apache/flink/blob/release-1.10/flink-yarn/src/main/java/org/apache/flink/yarn/cli/FlinkYarnSessionCli.java#L899_ --- The third-party dependencies might utilize this key as well, as [~lzljs3620320] mentioned. My intention is similar to how [~xymaqingxiang] mentioned that setting the value to a tmp directory under the working directory of Yarn container. was (Author: victor-wong): [~rongr] Thanks for your reply! _Could you elaborate what does it mean by Flink-YARN default the value to /tmp? I am guessing you mean JVM default the value to /tmp ? _ --- Yes, I mean that Flink-YARN is using the default value of JVM, which is "/tmp". _So far the only place I can see in flink yarn code utilizing this key is: https://github.com/apache/flink/blob/release-1.10/flink-yarn/src/main/java/org/apache/flink/yarn/cli/FlinkYarnSessionCli.java#L899 _ --- The third-party dependencies might utilize this key as well, as [~lzljs3620320] mentioned. My intention is similar to how [~xymaqingxiang] mentioned that setting the value to a tmp directory under the working directory of Yarn container. > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Comment Edited] (FLINK-15447) Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp"
[ https://issues.apache.org/jira/browse/FLINK-15447?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17018462#comment-17018462 ] Victor Wong edited comment on FLINK-15447 at 1/18/20 1:40 AM: -- [~rongr] Thanks for your reply! _Could you elaborate what does it mean by Flink-YARN default the value to /tmp? I am guessing you mean JVM default the value to /tmp ?_ --- Yes, I mean that Flink-YARN is using the default value of JVM, which is "/tmp". _So far the only place I can see in flink yarn code utilizing this key is: https://github.com/apache/flink/blob/release-1.10/flink-yarn/src/main/java/org/apache/flink/yarn/cli/FlinkYarnSessionCli.java#L899_ --- The third-party dependencies might utilize this key as well, as [~lzljs3620320] mentioned. My intention is similar to how [~xymaqingxiang] mentioned that setting the value to a tmp directory under the working directory of Yarn container. was (Author: victor-wong): [~rongr] Thanks for your reply! _Could you elaborate what does it mean by Flink-YARN default the value to /tmp? I am guessing you mean JVM default the value to /tmp ?_ --- Yes, I mean that Flink-YARN is using the default value of JVM, which is "/tmp". _So far the only place I can see in flink yarn code utilizing this key is: https://github.com/apache/flink/blob/release-1.10/flink-yarn/src/main/java/org/apache/flink/yarn/cli/FlinkYarnSessionCli.java#L899_ --- The third-party dependencies might utilize this key as well, as [~lzljs3620320] mentioned. My intention is similar to how [~xymaqingxiang] mentioned that setting the value to a tmp directory under the working directory of Yarn container. > Change "java.io.tmpdir" of JM/TM on Yarn to "{{PWD}}/tmp" > --- > > Key: FLINK-15447 > URL: https://issues.apache.org/jira/browse/FLINK-15447 > Project: Flink > Issue Type: Improvement > Components: Deployment / YARN >Affects Versions: 1.9.1 >Reporter: Victor Wong >Priority: Major > > Currently, when running Flink on Yarn, the "java.io.tmpdir" property is set > to the default value, which is "/tmp". > > Sometimes we ran into exceptions caused by a full "/tmp" directory, which > would not be cleaned automatically after applications finished. > I think we can set "java.io.tmpdir" to "PWD/tmp" directory, or > something similar. "PWD" will be replaced with the true working > directory of JM/TM by Yarn, which will be cleaned automatically. > -- This message was sent by Atlassian Jira (v8.3.4#803005)