[GitHub] flink issue #3151: [FLINK-4364][runtime]mplement TaskManager side of heartbe...
Github user wangzhijiang999 commented on the issue: https://github.com/apache/flink/pull/3151 @tillrohrmann , I have submitted the updates that may cover your suggestions. There are still two issues that I am not confirmed. First, for heartbeat interval and timeout default values in **ConfigConstants**, they are not invalid currently and you can modify it with your professional experience. Second, the introduction of **ScheduledExecutorService** in **RPCService**, my initial idea is trying to use the existing scheduler in **RPCService**, but it can not be got from **AkkaRPCService** implementation. Another way is to replace the current **ScheduledExecutorService** parameter with **RPCService** in construction of **HeartbeatManagerSenderImpl**, and the **RPCService** can also schedule the heartbeat request. But the return value of **scheduleRunnable** method in **RPCService** is conflict with that in **HeartbeatManagerSenderImpl**. So I just bring another single thread pool in **RPCService** for use currently. Maybe the number of threads in pool can refer to number of cpu cores. Maybe there are still something to be polished, and I am willing for further modifications by your comments. BTW, the heartbeat interaction between TM and RM will be submitted in another PR after this confirmation because of some common points. --- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---
[GitHub] flink issue #3151: [FLINK-4364][runtime]mplement TaskManager side of heartbe...
Github user wangzhijiang999 commented on the issue: https://github.com/apache/flink/pull/3151 @tillrohrmann , I am preparing the testing code and can submit the updates this week. Thank you for continuous help! --- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---
[GitHub] flink issue #3151: [FLINK-4364][runtime]mplement TaskManager side of heartbe...
Github user tillrohrmann commented on the issue: https://github.com/apache/flink/pull/3151 Thanks @wangzhijiang999. Ping me once you've updated the PR. I think it would make sense to test that the respective steps are taken if the JM detects that a TM died and vice versa. For that you basically only need to start a JM and a TM and let them connect. --- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---
[GitHub] flink issue #3151: [FLINK-4364][runtime]mplement TaskManager side of heartbe...
Github user wangzhijiang999 commented on the issue: https://github.com/apache/flink/pull/3151 @tillrohrmann . Thank you for detail review and comments! This PR just submit the heartbeat logic in TM side, because there is already a jira of JM heartbeat side. For my implementation, the JM initiates the heartbeat with **HeartbeatManagerSenderImpl** and the TM responses the heartbeat with **HeartbeatManagerImpl**. So the heartbeat process is one-way. I think it is better to submit the JM heartbeat logic in this PR in order to understand easily. I will modify this PR soon, and for testing there already exists the UT for basic heartbeat logic. Do you mean to add some ITCases? --- If your project is set up for it, you can reply to this email and have your reply appear on GitHub as well. If your project does not have this feature enabled and wishes so, or if the feature is enabled but not working, please contact infrastructure at infrastruct...@apache.org or file a JIRA ticket with INFRA. ---
