[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17776381#comment-17776381 ] Matthias J. Sax commented on KAFKA-6520: Could we use Admin.describeCluster() which returns the list of brokers that are currently online, and go into DISCONNECTED if we don't get anything back? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Vince Mu >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17537860#comment-17537860 ] Matthias J. Sax commented on KAFKA-6520: It's still an open item – and it's complex (for details, see the KIP discussion thread). > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Vince Mu >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.20.7#820007)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17537647#comment-17537647 ] Karthik Raman commented on KAFKA-6520: -- [~mjsax] / [~Yohan123] : Just wondering if there is any update on this issue to be added in any latest releases? Thank you. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Vince Mu >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.20.7#820007)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17273249#comment-17273249 ] Matthias J. Sax commented on KAFKA-6520: i am wondering if https://issues.apache.org/jira/browse/KAFKA-10866 (merge recently) is something we could exploit to implement a DISCONNECT state? The new metadata contains a `receivedTimestamp` field and thus we could track the time difference of "now" and the last received fetch response. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Vince Mu >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17246254#comment-17246254 ] Walker Carlson commented on KAFKA-6520: --- [https://cwiki.apache.org/confluence/x/lCvZCQ] changes the status for Streams. It makes the ERROR status terminal. Also the new handler added to Streams in KIP-671 (https://cwiki.apache.org/confluence/x/lkN4CQ) should transit such exceptions to state ERROR. This seems relevant to this issue and may change what is need for the PR. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Vince Mu >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17150519#comment-17150519 ] Boyang Chen commented on KAFKA-6520: [~VinceMu] Gave you contributor permission and assign the ticket to you. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Vince Mu >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17149490#comment-17149490 ] Vince Mu commented on KAFKA-6520: - Created a work in progress PR. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17121035#comment-17121035 ] Vince Mu commented on KAFKA-6520: - My approach for this would be to create a new metric, fetch-disconnect-rate, which would represent the number of disconnects over a short window of time. Disconnects could be recorded in the onFailure handler for the fetches being sent. In the Kafka streams container or in the streamThread we could read this metric and set the state to DISCONNECTED if the rate exceeds a certain threshold. However, I question the need for an entirely new state if it only serves to inform the user of connectivity. Would simply exposing a new metric be enough to achieve this? Appreciate any thoughts on this matter. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17117161#comment-17117161 ] Matthias J. Sax commented on KAFKA-6520: Would be great if [~guozhang] and/or [~hachikuji] could chime in. I am not totally familiar with all the details in the consumer code base. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17115097#comment-17115097 ] Vince Mu commented on KAFKA-6520: - [~mjsax] your explanation about coordinator metadata and liveness makes perfect sense. Thanks for that. Regarding the implementation of a disconnected timeout, I'm not sure whether introducing a disconnected timeout and measuring the timeout of each fetch request is necessary. It seems like the ConsumerNetworkClient and NetworkClient will already fail an unsent or transmitted fetch request with a disconnected exception it's connection to a node dies. So instead of throwing a disconnected exception based on whether all fetch requests timeout, we could instead throw a disconnected exception if all the fetch requests fail with a disconnected exception. I feel like this might be a simpler solution that uses what is already there. Thoughts on this? Please feel free to correct me. I'm still learning the code base bit by bit. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17112676#comment-17112676 ] Matthias J. Sax commented on KAFKA-6520: The consumer only stores/caches the last known coordinator metadata. There is not coordinator liveness check. The consumer only talks to the coordinator during a rebalance. However, if a network issue occurs during regular processing and the consumer cannot fetch data from the brokers, the cached coordinator metadata stays the same (ie, the coordinator is still known). Does this make sense? In the end, I believe without a consumer change we cannot really resolve the issue though. Maybe we could introduce a new "disconnected timeout" to the consumer: internally, the consumer sends fetch requests on a regular basis. Those fetch request would timeout if there is a network issue. Currently, the consumer "swallows" those timeout exceptions and just keeps retrying. `poll()` would just return zero records, but never rethrow a TimeoutException. The new "disconnected timeout" could be used to set a limit how long fetch request should be retried: if all fetch requests timeout for a period longer than "disconnected timeout", poll() could throw a "DisconnectedException" in the next `poll()` call (to reuse TimeoutException could be miss leading?). By default, the new "disconnected timeout" would be set to MAX_VALUE, and thus the default behavior would not change. Within KafkaStreams, we can set this new timeout to a smaller value and catch the exception (and change the Kafka Streams state to DISCONNECT). On a consecutive `poll()` that does not throw, we set the state back to RUNNING. Thoughts? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17112118#comment-17112118 ] Vince Mu commented on KAFKA-6520: - I think the easiest and perhaps simplest way is to check if the stream client is connected is to poll for whether the group coordinator is known in the stream thread's consumer. However this would involve changes to the consumer api. However [~hachikuji] mentioned earlier in the [PR |[https://github.com/apache/kafka/pull/6594]], that coordinator liveliness is not enough to determine whether the streams client is disconnnected. Why is the case? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17110830#comment-17110830 ] Matthias J. Sax commented on KAFKA-6520: [~guozhang] I think your idea about leveraging KIP-572 might not work. I dug though the code and none of the blocking calls that might through a `TimeoutException` are on the regular processing code path. Only during task initialization or restore, blocking calls are made. During normal processing, only `poll()` / `pause()` / `resume()` are called and those methods don't throw a `TimeoutException`. Thoughts? [~VinceMu] Yes, the main purpose is to have a KafkaStreams client state DISCONNECTED. Thread state is an internal implementation detail. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17110746#comment-17110746 ] Vince Mu commented on KAFKA-6520: - [~guozhang] that makes sense. I just want to clarify whether we are adding a DISCONNECTED state purely to streamThread or whether we need to add it KafkaStreams client as well? If it's the latter, would it make sense to transit the client to the DISCONNECTED state when all it's streamThreads are in the DISCONNECTED state? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17110447#comment-17110447 ] Guozhang Wang commented on KAFKA-6520: -- Just to throw some more ideas here: the embedded clients are now all using async network IO and hence one would never get ClosedChannelException, instead, they will eventually get TimeoutException when the broker is actually offline. [~mjsax] is currently working on KIP-572 to let Streams be more resilient to such connectivity issues (broker unavailable is exposed the same as network in-connectivity), while if we have N tasks, we would still continue when only a subset of them cannot progress. On the other hand, a Streams client may talk to multiple brokers on behalf of different tasks, and as long as one of the tasks can still progress -- meaning, its corresponding required brokers are still reachable -- then we would not need to mark the client as disconnected. Following this train of thoughts, I feel that we would only transit to the DISCONNECTED state if none of the tasks are progressing, indicating that none of the required brokers are available at the moment. Does that make sense? If yes then the scope of it can be much simplified, and maybe we can also just piggyback the proposal as part of KIP-572 so that we do not need a separate KIP. Of course, implementation wise Vince and Matthias can still proceed independently. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17109807#comment-17109807 ] Vince Mu commented on KAFKA-6520: - The KIP details the adding of a "DISCONNECTED" state to StreamThread while this ticket seems to concerned with the state of the KafkaStreams connection. Should we also be adding a "DICONNECTED" state to the KafkaStreams connection? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17107917#comment-17107917 ] Matthias J. Sax commented on KAFKA-6520: This is a quite tricky one... You can of course try. Maybe study the previous PR and the discussion on it to get some context before starting to write code? – And of course, feel free to ask any questions. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=17107900#comment-17107900 ] Vince Mu commented on KAFKA-6520: - Can i be assigned this ticket? I would love to give it a shot. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16981234#comment-16981234 ] Matthias J. Sax commented on KAFKA-6520: Unassigned this ticket due to inactivity. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16981233#comment-16981233 ] ASF GitHub Bot commented on KAFKA-6520: --- mjsax commented on pull request #6594: [KAFKA-6520] Add DISCONNECTED state to Kafka Streams URL: https://github.com/apache/kafka/pull/6594 This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > KIP WIP: > [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. This is a link to a related > issue. > - > Update: there are some discussions on the PR itself which leads me to think > that a more general solution should be at the ClusterConnectionStates rather > than at the Streams or even Consumer level. One proposal would be: > * Add a new metric named `failedConnection` in SelectorMetrics which is > recorded at `connect()` and `pollSelectionKeys()` functions, upon capture the > IOException / RuntimeException which indicates the connection disconnected. > * And then users of Consumer / Streams can monitor on this metric, which > normally will only have close to zero values as we have transient > disconnects, if it is spiking it means the brokers are consistently being > unavailable indicting the state. > [~Yohan123] WDYT? -- This message was sent by Atlassian Jira (v8.3.4#803005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16882573#comment-16882573 ] Brian Putt commented on KAFKA-6520: --- Any updates on this? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16820488#comment-16820488 ] ASF GitHub Bot commented on KAFKA-6520: --- ConcurrencyPractitioner commented on pull request #6594: [KAFKA-6520] Add DISCONNECTED state to Kafka Streams URL: https://github.com/apache/kafka/pull/6594 We wish to add a DISCONNECTED state to Kafka Streams. Please see KIP-457 for details. https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams ### Committer Checklist (excluded from commit message) - [ ] Verify design and implementation - [ ] Verify test coverage and CI build status - [ ] Verify documentation (including upgrade notes) This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. For queries about this service, please contact Infrastructure at: us...@infra.apache.org > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16819258#comment-16819258 ] Richard Yu commented on KAFKA-6520: --- Hello, I've created a KIP for this issue so that we could get some progress. Heres the link :) : [https://cwiki.apache.org/confluence/display/KAFKA/KIP-457%3A+Add+DISCONNECTED+status+to+Kafka+Streams] > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16556958#comment-16556958 ] Matthias J. Sax commented on KAFKA-6520: No worries. All good :) A KIP is a "Kafka Improvement Proposal": [https://cwiki.apache.org/confluence/display/KAFKA/Kafka+Improvement+Proposals] It's only required for major features and public API changes. This ticket does not need a KIP. Thus, you should not start a DISCUSS thread and voting is not required either. It might just be helpful to do a similar writeup (could be a google doc) for the design – might help to get "your feet wet" as you phrased it. If you don't want to do it, it's fine, too. I recommend it, as it seems to be more efficient if you do a full design instead and ask all open question at once if you have any instead of individual questions going forth and back in the comment section on the ticket. Let us know if you need any help. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16556767#comment-16556767 ] Harjit Singh commented on KAFKA-6520: - Pardon my ignorance. I'mm trying to get my feet wet with the process. What's KIP? Also, I was trying to read some docs from the docs folder but can't get them to open in the browser. Do you know how can I open them? Right now when Open in a browser, I see a blank page. I need to read some more on the design of Kafka before I can submit my design. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16556757#comment-16556757 ] Matthias J. Sax commented on KAFKA-6520: [~harjitdotsingh] If you want to work on this ticket, it's your call to make a proposal :) If you want to make a full design before opening a PR, please prepare a design document (similar to a KIP and we can discuss it). Otherwise, you can also just open a PR and we discuss your proposal there during review. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16556479#comment-16556479 ] Harjit Singh commented on KAFKA-6520: - I didn't see the proposed options. I have figured out where the State Change takes place in the streams. Are we saying that we need to introduce a new state ? Right now the valid states are documented below So if we are planning to introduce a new State, DISCONNECTED, what would be the valid transitions for it ? REBALANCING,RUNNING, PENDING_SHUTDOWN ? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[
https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16554817#comment-16554817
]
Guozhang Wang commented on KAFKA-6520:
--
Just want to re-emphasize that, it is not only a Streams feature request: if
you start a Java consumer today talking to a broker, and the broker is not
available, that consumer client would not throw an exception either. This was
by-design of the consumer at that time:
{code}
When KafkaConsumer.poll is called, in which updateAssignmentMetadataIfNeeded is
called and coordinator.poll() is called. If the broker(s) are not available
then the last call will always return false after timeout, and hence
KafkaConsumer.poll will return empty set without any indicator that brokers are
disconnect. In fact, if the brokers are unavailable at the very beginning when
streams application starts, we will hit the same issue as the instance will
also be in the RUNNING state since the onPartitionRevoked would not be called
ever.
...
On the other hand, it is by-design of the consumer to not let users worry about
the connectivity of the broker, or to handle any disconnect issues. Here is my
proposed options:
{code}
> When a Kafka Stream can't communicate with the server, it's Status stays
> RUNNING
>
>
> Key: KAFKA-6520
> URL: https://issues.apache.org/jira/browse/KAFKA-6520
> Project: Kafka
> Issue Type: Improvement
> Components: streams
>Reporter: Michael Kohout
>Assignee: Milind Jain
>Priority: Major
> Labels: newbie, user-experience
>
> When you execute the following scenario the application is always in RUNNING
> state
>
> 1)start kafka
> 2)start app, app connects to kafka and starts processing
> 3)kill kafka(stop docker container)
> 4)the application doesn't give any indication that it's no longer
> connected(Stream State is still RUNNING, and the uncaught exception handler
> isn't invoked)
>
>
> It would be useful if the Stream State had a DISCONNECTED status.
>
> See
> [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ]
> for a discussion from the google user forum.
> [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a
> related issue.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16554771#comment-16554771 ] Matthias J. Sax commented on KAFKA-6520: As the ticket description above says: > It would be useful if the Stream State had a DISCONNECTED status. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16554739#comment-16554739 ] Harjit Singh commented on KAFKA-6520: - Should the stream stop running or log an error ? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16554730#comment-16554730 ] Matthias J. Sax commented on KAFKA-6520: [~harjitdotsingh] Well. That is what this ticket is about, isn't it? It doesn't matter if brokers are not reachable on startup or go down while the application is running. For both cases, it just tries to reconnect to the brokers silently. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16554726#comment-16554726 ] Harjit Singh commented on KAFKA-6520: - Even if I don't run Kafka the wordCountDemo program doesn't complain not being able to connect to the broker. Shouldn't we be seeing the error on startup ? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16553124#comment-16553124 ] Guozhang Wang commented on KAFKA-6520: -- [~milindjain] I'm not very familiar about stormKafka, maybe they are still using the old scala client that has a different behavior. When you tried console consumer, did you used new consumer (setting --new-consumer) or not? If you are using new consumer, then even when the brokers become unavailable it should not be throwing errors and die. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16551089#comment-16551089 ] Matthias J. Sax commented on KAFKA-6520: As mentioned, try the `WordCountExample` that is showcased in the quickstart. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16551025#comment-16551025 ] Harjit Singh commented on KAFKA-6520: - [~mjsax] - Is there a sample app for Streaming which I can use or do I have to write code ? I'm familiar with Kafka, was looking for a sample for streaming so that I can debug this issue. Thanks - Harjit > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16551012#comment-16551012 ] Matthias J. Sax commented on KAFKA-6520: [~harjitdotsingh] Not sure what you exactly mean... To write/read to/from Kafka, you can use command line tools `bin/kafka-console-producer` and `bin/kafka-console-consumer`. There is also a quickstart example application for Kafka Streams. I would recommend to walk through the Kafka quickstart that show cases all of them: https://kafka.apache.org/documentation/#quickstart > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16550163#comment-16550163 ] Harjit Singh commented on KAFKA-6520: - Which App can I use to connect to Kafka? Is there a sample app available anywhere? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16544423#comment-16544423 ] Milind Jain commented on KAFKA-6520: Hi [~guozhang], Sorry about the delay, What I have seen is that the kafka console consumer and stormkafka consumer throws error when the broker is not available, where as the Kafka stream doesn't give any warning or error According to you is similar behavior observed in case of Kafka consumer, console consumer? What do you see when you kill the broker when consumer is still running? Regards > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[
https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16544335#comment-16544335
]
Guozhang Wang commented on KAFKA-6520:
--
Here is my current analysis on this issue:
When {{KafkaConsumer.poll}} is called, in which
{{updateAssignmentMetadataIfNeeded}} is called and {{coordinator.poll()}} is
called. If the broker(s) are not available then the last call will always
return false after timeout, and hence {{KafkaConsumer.poll}} will return empty
set without any indicator that brokers are disconnect. In fact, if the brokers
are unavailable at the very beginning when streams application starts, we will
hit the same issue as the instance will also be in the {{RUNNING}} state since
the onPartitionRevoked would not be called ever.
So I think if we want to improve on this situation it should be on the consumer
client, not the streams client. I.e. today if you use a KafkaConsumer to fetch
data directly, and the brokers become unavailable, you will observe the same
scenario. On the other hand, it is by-design of the consumer to not let users
worry about the connectivity of the broker, or to handle any disconnect issues.
Here is my proposed options:
1. Expose metrics from Consumer#NetworkClient on the
{{ClusterConnectionStates}}. And then let users to watch on this metrics to
watch out for dis-connectivity issues.
2. We can add a new State, say named {{IDLE}}, to StreamThread, that if
{{consumer.poll}} returns nothing, and in the same iteration, restore consumer
also returns nothing, transit to {{IDLE}}. So only {{RUNNING}} can be transited
to {{IDLE}} and {{IDLE}} can only be transited to {{RUNNING}} (if the next poll
call returns some data) or {{PARTITION_REVOKED}} (if the next poll call
triggers rebalance, and the callback is triggered). For global thread it is
simpler, that if one poll call returns nothing transit to {{IDLE}}. And also
the KafkaStreams can add a new {{IDLE}} as well, which will be transited from
{{RUNNING}} if all its thread (including global thread) transit to {{IDLE}}.
Thoughts?
> When a Kafka Stream can't communicate with the server, it's Status stays
> RUNNING
>
>
> Key: KAFKA-6520
> URL: https://issues.apache.org/jira/browse/KAFKA-6520
> Project: Kafka
> Issue Type: Improvement
> Components: streams
>Reporter: Michael Kohout
>Assignee: Milind Jain
>Priority: Major
> Labels: newbie, user-experience
>
> When you execute the following scenario the application is always in RUNNING
> state
>
> 1)start kafka
> 2)start app, app connects to kafka and starts processing
> 3)kill kafka(stop docker container)
> 4)the application doesn't give any indication that it's no longer
> connected(Stream State is still RUNNING, and the uncaught exception handler
> isn't invoked)
>
>
> It would be useful if the Stream State had a DISCONNECTED status.
>
> See
> [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ]
> for a discussion from the google user forum.
> [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a
> related issue.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16524331#comment-16524331 ] Guozhang Wang commented on KAFKA-6520: -- Hi [~milindjain], are you still working on this ticket? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16462301#comment-16462301 ] Matthias J. Sax commented on KAFKA-6520: If you can reproduce the issue as describe, the next step would be to think how it could be fixed, ie, how can we detect this case to transit into a new state. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16460076#comment-16460076 ] Milind Jain commented on KAFKA-6520: Hi, I tried reproducing the issue with /kafka/streams/examples/src/main/java/org/apache/kafka/streams/examples/wordcount/WordCountDemo.java On stopping the broker when WordCountDemo is still running, I am not able to see any error in the WordCountDemo, It keeps running as if the broker is still alive. [~mjsax] [~mwkohout] What should I do next. I am able to reproduce the issue. > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[
https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16404326#comment-16404326
]
Matthias J. Sax commented on KAFKA-6520:
This issue is target to Kafka's Streams API, ie, if one starts a
{{KafkaStreams}} instance. Check the docs for details about Streams API:
[https://kafka.apache.org/10/documentation/streams/] Hope this helps.
> When a Kafka Stream can't communicate with the server, it's Status stays
> RUNNING
>
>
> Key: KAFKA-6520
> URL: https://issues.apache.org/jira/browse/KAFKA-6520
> Project: Kafka
> Issue Type: Improvement
> Components: streams
>Reporter: Michael Kohout
>Assignee: Milind Jain
>Priority: Major
> Labels: newbie, user-experience
>
> When you execute the following scenario the application is always in RUNNING
> state
>
> 1)start kafka
> 2)start app, app connects to kafka and starts processing
> 3)kill kafka(stop docker container)
> 4)the application doesn't give any indication that it's no longer
> connected(Stream State is still RUNNING, and the uncaught exception handler
> isn't invoked)
>
>
> It would be useful if the Stream State had a DISCONNECTED status.
>
> See
> [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ]
> for a discussion from the google user forum.
> [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a
> related issue.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[
https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16404180#comment-16404180
]
Milind Jain commented on KAFKA-6520:
Hi [~mjsax] [~mwkohout]
I tried reproducing the issue
On using Storm kafka Spout I am getting the below exception on stopping the
broker
{color:#FF}Caused by: java.nio.channels.ClosedChannelException{color}
{color:#FF} at
kafka.network.BlockingChannel.send(BlockingChannel.scala:100)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer.liftedTree1$1(SimpleConsumer.scala:78)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer.kafka$consumer$SimpleConsumer$$sendRequest(SimpleConsumer.scala:68)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer$$anonfun$fetch$1$$anonfun$apply$mcV$sp$1.apply$mcV$sp(SimpleConsumer.scala:112)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer$$anonfun$fetch$1$$anonfun$apply$mcV$sp$1.apply(SimpleConsumer.scala:112)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer$$anonfun$fetch$1$$anonfun$apply$mcV$sp$1.apply(SimpleConsumer.scala:112)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at kafka.metrics.KafkaTimer.time(KafkaTimer.scala:33)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer$$anonfun$fetch$1.apply$mcV$sp(SimpleConsumer.scala:111)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer$$anonfun$fetch$1.apply(SimpleConsumer.scala:111)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer$$anonfun$fetch$1.apply(SimpleConsumer.scala:111)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at kafka.metrics.KafkaTimer.time(KafkaTimer.scala:33)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.consumer.SimpleConsumer.fetch(SimpleConsumer.scala:110)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
kafka.javaapi.consumer.SimpleConsumer.fetch(SimpleConsumer.scala:47)
~[kafka_2.10-0.8.2.0.jar:?]{color}
{color:#FF} at
org.apache.storm.kafka.KafkaUtils.fetchMessages(KafkaUtils.java:191)
~[storm-kafka-1.0.3.jar:1.0.3]{color}
And in console consumer/producer I am getting
{color:#f6c342}WARN Connection to node 0 could not be established. Broker may
not be available. (org.apache.kafka.clients.NetworkClient){color}
So what should I try doing next?
> When a Kafka Stream can't communicate with the server, it's Status stays
> RUNNING
>
>
> Key: KAFKA-6520
> URL: https://issues.apache.org/jira/browse/KAFKA-6520
> Project: Kafka
> Issue Type: Improvement
> Components: streams
>Reporter: Michael Kohout
>Assignee: Milind Jain
>Priority: Major
> Labels: newbie, user-experience
>
> When you execute the following scenario the application is always in RUNNING
> state
>
> 1)start kafka
> 2)start app, app connects to kafka and starts processing
> 3)kill kafka(stop docker container)
> 4)the application doesn't give any indication that it's no longer
> connected(Stream State is still RUNNING, and the uncaught exception handler
> isn't invoked)
>
>
> It would be useful if the Stream State had a DISCONNECTED status.
>
> See
> [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ]
> for a discussion from the google user forum.
> [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a
> related issue.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16394665#comment-16394665 ] Matthias J. Sax commented on KAFKA-6520: Exactly. Kafka servers are called brokers. :) > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16394656#comment-16394656 ] Milind Jain commented on KAFKA-6520: By broker you mean stopping the kafka-server while the consumer is still running and check the behaviour of the consumer? > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Assignee: Milind Jain >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[
https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16394652#comment-16394652
]
Matthias J. Sax commented on KAFKA-6520:
[~milindjain], welcome to the community and thanks for you interest in
contributing to Kafka!
I added you to the list of contributors. This allows you to assign JIRAs to
yourself. So feel free to grab any unassigned Jira you want to work on.
About the ticket itself, I never thought how this could be resolved. Maybe it's
a good start to try to reproduce the issue locally by running a broker,
starting a Kafka Streams application in you IDE, stop the broker to figure out
how the app behaves. A good starting point would to look into
{{StreamThead#run()}} method. Also, check out this wiki page that describes the
architecture of the library to some extend:
[https://cwiki.apache.org/confluence/display/KAFKA/Kafka+Streams+Architecture]
If you are completely new to Kafka, you should familiarize yourself with the
basic abstraction of brokers, topic, partitions, clients, consumer groups,
group rebalance first. There is tons of material online. Let us know if you
need any other help to get started.
Also check out [https://kafka.apache.org/contributing]
> When a Kafka Stream can't communicate with the server, it's Status stays
> RUNNING
>
>
> Key: KAFKA-6520
> URL: https://issues.apache.org/jira/browse/KAFKA-6520
> Project: Kafka
> Issue Type: Improvement
> Components: streams
>Reporter: Michael Kohout
>Assignee: Milind Jain
>Priority: Major
> Labels: newbie, user-experience
>
> When you execute the following scenario the application is always in RUNNING
> state
>
> 1)start kafka
> 2)start app, app connects to kafka and starts processing
> 3)kill kafka(stop docker container)
> 4)the application doesn't give any indication that it's no longer
> connected(Stream State is still RUNNING, and the uncaught exception handler
> isn't invoked)
>
>
> It would be useful if the Stream State had a DISCONNECTED status.
>
> See
> [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ]
> for a discussion from the google user forum.
> [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a
> related issue.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
[jira] [Commented] (KAFKA-6520) When a Kafka Stream can't communicate with the server, it's Status stays RUNNING
[ https://issues.apache.org/jira/browse/KAFKA-6520?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=16394641#comment-16394641 ] Milind Jain commented on KAFKA-6520: I am new to kafka. I would like to work on this issue. [~mwkohout] Can you please help me in getting started? Thanks Milind > When a Kafka Stream can't communicate with the server, it's Status stays > RUNNING > > > Key: KAFKA-6520 > URL: https://issues.apache.org/jira/browse/KAFKA-6520 > Project: Kafka > Issue Type: Improvement > Components: streams >Reporter: Michael Kohout >Priority: Major > Labels: newbie, user-experience > > When you execute the following scenario the application is always in RUNNING > state > > 1)start kafka > 2)start app, app connects to kafka and starts processing > 3)kill kafka(stop docker container) > 4)the application doesn't give any indication that it's no longer > connected(Stream State is still RUNNING, and the uncaught exception handler > isn't invoked) > > > It would be useful if the Stream State had a DISCONNECTED status. > > See > [this|https://groups.google.com/forum/#!topic/confluent-platform/nQh2ohgdrIQ] > for a discussion from the google user forum. > [This|https://issues.apache.org/jira/browse/KAFKA-4564] is a link to a > related issue. -- This message was sent by Atlassian JIRA (v7.6.3#76005)
