[
https://issues.apache.org/jira/browse/KAFKA-19485?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Shivsundar R updated KAFKA-19485:
---------------------------------
Description:
There are some race scenarios in multi-broker environments where
acknowledgements are sent on an initial epoch in a ShareSession.
This could arise if
# Consumer subscribed to a partition whose leader was node-0.
# Broker restart happens and node-0 is elected leader again. Broker starts a
new ShareSession.
# Background thread sends a fetch request with non-zero epoch.
# Broker responds with SHARE_SESSION_NOT_FOUND.
# Client updates session epoch to 0 once it receives this error.
# Application thread processing the previous fetch, completes and sends acks
to piggyback on next fetch.
# Consumer is no longer *subscribed* to this partition (from heartbeat
response) due to rebalance/change in assignment.
# Next fetch will send the piggyback acknowledgements on the fetch for
previously subscribed partitions resulting in error from broker
("Acknowledgments sent on initial epoch").
*Fix :* Add a check before sending out acknowledgments if we are on an initial
epoch and add necessary tests.
was:
There are some race scenarios in multi-broker environments where
acknowledgements are sent on an initial epoch in a ShareSession.
This could arise if
# Consumer subscribed to a partition whose leader was node-0.
# Broker restart happens and node-0 is elected leader again. Broker starts a
new ShareSession.
# Background thread sends a fetch request with non-zero epoch.
# Broker responds with SHARE_SESSION_NOT_FOUND.
# Client updates session epoch to 0 once it receives this error.
# Application thread processing the previous fetch, completes and sends acks
to piggyback on next fetch.
# Next fetch will send the piggyback acknowledgements on the fetch resulting
in error from broker ("Acknowledgments sent on initial epoch").
*Fix :* Add a check before sending out acknowledgments if we are on an initial
epoch.
> Acknowledgements should not be sent on initial epoch of ShareFetch
> ------------------------------------------------------------------
>
> Key: KAFKA-19485
> URL: https://issues.apache.org/jira/browse/KAFKA-19485
> Project: Kafka
> Issue Type: Sub-task
> Reporter: Shivsundar R
> Assignee: Shivsundar R
> Priority: Major
>
> There are some race scenarios in multi-broker environments where
> acknowledgements are sent on an initial epoch in a ShareSession.
> This could arise if
> # Consumer subscribed to a partition whose leader was node-0.
> # Broker restart happens and node-0 is elected leader again. Broker starts a
> new ShareSession.
> # Background thread sends a fetch request with non-zero epoch.
> # Broker responds with SHARE_SESSION_NOT_FOUND.
> # Client updates session epoch to 0 once it receives this error.
> # Application thread processing the previous fetch, completes and sends acks
> to piggyback on next fetch.
> # Consumer is no longer *subscribed* to this partition (from heartbeat
> response) due to rebalance/change in assignment.
> # Next fetch will send the piggyback acknowledgements on the fetch for
> previously subscribed partitions resulting in error from broker
> ("Acknowledgments sent on initial epoch").
> *Fix :* Add a check before sending out acknowledgments if we are on an
> initial epoch and add necessary tests.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
