[jira] [Commented] (KAFKA-14379) consumer should refresh preferred read replica on update metadata
[ https://issues.apache.org/jira/browse/KAFKA-14379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17632031#comment-17632031 ] Deng Ziming commented on KAFKA-14379: - Hello [~jeffkbkim] , you created another duplicates of this issue, I have closed them, plz not create identical issues. > consumer should refresh preferred read replica on update metadata > - > > Key: KAFKA-14379 > URL: https://issues.apache.org/jira/browse/KAFKA-14379 > Project: Kafka > Issue Type: Bug >Reporter: Jeff Kim >Assignee: Jeff Kim >Priority: Major > > The consumer (fetcher) refreshes the preferred read replica only on three > conditions: > # the consumer receives an OFFSET_OUT_OF_RANGE error > # the follower does not exist in the client's metadata (i.e., offline) > # after metadata.max.age.ms (5 min default) > For other errors, it will continue to reach to the possibly unavailable > follower and only after 5 minutes will it refresh the preferred read replica > and go back to the leader. > A specific example is when a partition is reassigned. the consumer will get > NOT_LEADER_OR_FOLLOWER which triggers a metadata update but the preferred > read replica will not be refreshed as the follower is still online. it will > continue to reach out to the old follower until the preferred read replica > expires. > the consumer can instead refresh its preferred read replica whenever it makes > a metadata update request. so when the consumer receives i.e. > NOT_LEADER_OR_FOLLOWER it can find the new preferred read replica without > waiting for the expiration. -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Commented] (KAFKA-14379) consumer should refresh preferred read replica on update metadata
[ https://issues.apache.org/jira/browse/KAFKA-14379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17634565#comment-17634565 ] Jeff Kim commented on KAFKA-14379: -- sorry for that. got an error when i clicked on create > consumer should refresh preferred read replica on update metadata > - > > Key: KAFKA-14379 > URL: https://issues.apache.org/jira/browse/KAFKA-14379 > Project: Kafka > Issue Type: Bug >Reporter: Jeff Kim >Assignee: Jeff Kim >Priority: Major > > The consumer (fetcher) refreshes the preferred read replica only on three > conditions: > # the consumer receives an OFFSET_OUT_OF_RANGE error > # the follower does not exist in the client's metadata (i.e., offline) > # after metadata.max.age.ms (5 min default) > For other errors, it will continue to reach to the possibly unavailable > follower and only after 5 minutes will it refresh the preferred read replica > and go back to the leader. > A specific example is when a partition is reassigned. the consumer will get > NOT_LEADER_OR_FOLLOWER which triggers a metadata update but the preferred > read replica will not be refreshed as the follower is still online. it will > continue to reach out to the old follower until the preferred read replica > expires. > the consumer can instead refresh its preferred read replica whenever it makes > a metadata update request. so when the consumer receives i.e. > NOT_LEADER_OR_FOLLOWER it can find the new preferred read replica without > waiting for the expiration. -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Commented] (KAFKA-14379) consumer should refresh preferred read replica on update metadata
[ https://issues.apache.org/jira/browse/KAFKA-14379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17638188#comment-17638188 ] Luke Chen commented on KAFKA-14379: --- [~jeffkbkim] , we recently also encountered this issue, do you plan to submit PR to fix this issue before v3.4.0/v3.3.2? > consumer should refresh preferred read replica on update metadata > - > > Key: KAFKA-14379 > URL: https://issues.apache.org/jira/browse/KAFKA-14379 > Project: Kafka > Issue Type: Bug >Reporter: Jeff Kim >Assignee: Jeff Kim >Priority: Major > > The consumer (fetcher) refreshes the preferred read replica only on three > conditions: > # the consumer receives an OFFSET_OUT_OF_RANGE error > # the follower does not exist in the client's metadata (i.e., offline) > # after metadata.max.age.ms (5 min default) > For other errors, it will continue to reach to the possibly unavailable > follower and only after 5 minutes will it refresh the preferred read replica > and go back to the leader. > A specific example is when a partition is reassigned. the consumer will get > NOT_LEADER_OR_FOLLOWER which triggers a metadata update but the preferred > read replica will not be refreshed as the follower is still online. it will > continue to reach out to the old follower until the preferred read replica > expires. > the consumer can instead refresh its preferred read replica whenever it makes > a metadata update request. so when the consumer receives i.e. > NOT_LEADER_OR_FOLLOWER it can find the new preferred read replica without > waiting for the expiration. > -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Commented] (KAFKA-14379) consumer should refresh preferred read replica on update metadata
[ https://issues.apache.org/jira/browse/KAFKA-14379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17638191#comment-17638191 ] Luke Chen commented on KAFKA-14379: --- Oh, sorry, I saw you have a WIP PR opened. Thanks. > consumer should refresh preferred read replica on update metadata > - > > Key: KAFKA-14379 > URL: https://issues.apache.org/jira/browse/KAFKA-14379 > Project: Kafka > Issue Type: Bug >Reporter: Jeff Kim >Assignee: Jeff Kim >Priority: Major > > The consumer (fetcher) refreshes the preferred read replica only on three > conditions: > # the consumer receives an OFFSET_OUT_OF_RANGE error > # the follower does not exist in the client's metadata (i.e., offline) > # after metadata.max.age.ms (5 min default) > For other errors, it will continue to reach to the possibly unavailable > follower and only after 5 minutes will it refresh the preferred read replica > and go back to the leader. > A specific example is when a partition is reassigned. the consumer will get > NOT_LEADER_OR_FOLLOWER which triggers a metadata update but the preferred > read replica will not be refreshed as the follower is still online. it will > continue to reach out to the old follower until the preferred read replica > expires. > the consumer can instead refresh its preferred read replica whenever it makes > a metadata update request. so when the consumer receives i.e. > NOT_LEADER_OR_FOLLOWER it can find the new preferred read replica without > waiting for the expiration. > -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Commented] (KAFKA-14379) consumer should refresh preferred read replica on update metadata
[ https://issues.apache.org/jira/browse/KAFKA-14379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17640158#comment-17640158 ] Jeff Kim commented on KAFKA-14379: -- [~showuon] yes, that is the current plan. can you share what error the client faced and the result? > consumer should refresh preferred read replica on update metadata > - > > Key: KAFKA-14379 > URL: https://issues.apache.org/jira/browse/KAFKA-14379 > Project: Kafka > Issue Type: Bug >Reporter: Jeff Kim >Assignee: Jeff Kim >Priority: Major > > The consumer (fetcher) refreshes the preferred read replica only on three > conditions: > # the consumer receives an OFFSET_OUT_OF_RANGE error > # the follower does not exist in the client's metadata (i.e., offline) > # after metadata.max.age.ms (5 min default) > For other errors, it will continue to reach to the possibly unavailable > follower and only after 5 minutes will it refresh the preferred read replica > and go back to the leader. > A specific example is when a partition is reassigned. the consumer will get > NOT_LEADER_OR_FOLLOWER which triggers a metadata update but the preferred > read replica will not be refreshed as the follower is still online. it will > continue to reach out to the old follower until the preferred read replica > expires. > the consumer can instead refresh its preferred read replica whenever it makes > a metadata update request. so when the consumer receives i.e. > NOT_LEADER_OR_FOLLOWER it can find the new preferred read replica without > waiting for the expiration. > -- This message was sent by Atlassian Jira (v8.20.10#820010)
[jira] [Commented] (KAFKA-14379) consumer should refresh preferred read replica on update metadata
[ https://issues.apache.org/jira/browse/KAFKA-14379?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17640278#comment-17640278 ] Jeff Kim commented on KAFKA-14379: -- updated from draft to open PR. > consumer should refresh preferred read replica on update metadata > - > > Key: KAFKA-14379 > URL: https://issues.apache.org/jira/browse/KAFKA-14379 > Project: Kafka > Issue Type: Bug >Reporter: Jeff Kim >Assignee: Jeff Kim >Priority: Major > > The consumer (fetcher) refreshes the preferred read replica only on three > conditions: > # the consumer receives an OFFSET_OUT_OF_RANGE error > # the follower does not exist in the client's metadata (i.e., offline) > # after metadata.max.age.ms (5 min default) > For other errors, it will continue to reach to the possibly unavailable > follower and only after 5 minutes will it refresh the preferred read replica > and go back to the leader. > A specific example is when a partition is reassigned. the consumer will get > NOT_LEADER_OR_FOLLOWER which triggers a metadata update but the preferred > read replica will not be refreshed as the follower is still online. it will > continue to reach out to the old follower until the preferred read replica > expires. > the consumer can instead refresh its preferred read replica whenever it makes > a metadata update request. so when the consumer receives i.e. > NOT_LEADER_OR_FOLLOWER it can find the new preferred read replica without > waiting for the expiration. > -- This message was sent by Atlassian Jira (v8.20.10#820010)
