[
https://issues.apache.org/jira/browse/KAFKA-13804?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Luke Chen updated KAFKA-13804:
------------------------------
Description:
Currently, when the broker hit exception, or SIGINT signal received, we will
first log the exception (or the reason we started to shutdown) ex:
_INFO Terminating process due to signal SIGINT_
And then go through the shutdown process, to close all the resources (i.e.
threads, sockets, metrics...), and in the end, output:
_[KafkaServer id=0] shut down completed (kafka.server.KafkaServer)_
Sometimes, during closing the resources, we will also hit some exceptions, or
waiting for some other dependencies close, and output many more logs. When we
tried to find out why the broker is shutdown from the log, we have to go
through many lines of logs from the tail, and skip some unrelated exceptions,
and then identify the specific reason why the broker shutting down.
We could improve it by adding the reason why the broker is shutting down at the
end of log, something like this:
{code:java}
[2022-04-07 18:19:33,005] INFO Terminating process due to signal SIGINT
(org.apache.kafka.common.utils.LoggingSignalHandler)
[2022-04-07 18:19:33,007] INFO [KafkaServer id=0] shutting down
(kafka.server.KafkaServer)
[2022-04-07 18:19:33,008] INFO [KafkaServer id=0] Starting controlled shutdown
(kafka.server.KafkaServer)
[2022-04-07 18:19:33,016] INFO [KafkaServer id=0] Controlled shutdown request
returned successfully after 6ms (kafka.server.KafkaServer)
....
[2022-04-07 18:19:33,227] INFO Broker and topic stats closed
(kafka.server.BrokerTopicStats)
[2022-04-07 18:19:33,227] INFO App info kafka.server for 0 unregistered
(org.apache.kafka.common.utils.AppInfoParser)
[2022-04-07 18:19:33,227] INFO [KafkaServer id=0] shut down completed
(kafka.server.KafkaServer)
// new added log
[2022-04-07 18:19:33,227] INFO [KafkaServer id=0] the broker is shut down due
to: signal SIGINT{code}
was:
Currently, when the broker hit exception, or SIGINT signal received, we will
first log the exception (or the reason we started to shutdown) ex:
_INFO Terminating process due to signal SIGINT_
And then go through the shutdown process, to close all the resources (i.e.
threads, sockets, metrics...), and in the end, output:
_[KafkaServer id=0] shut down completed (kafka.server.KafkaServer)_
Sometimes, during closing the resources, we will also hit some exceptions, or
waiting for some other dependencies close, and output many more logs. In the
end, when we tried to find out why the broker is shutdown, we have to go
through many lines of logs from the tail, and identify the specific reason why
the broker shutting down.
We could improve it by adding the reason why the broker is shutting down at the
end of log, something like this:
{code:java}
[2022-04-07 18:19:33,005] INFO Terminating process due to signal SIGINT
(org.apache.kafka.common.utils.LoggingSignalHandler)
[2022-04-07 18:19:33,007] INFO [KafkaServer id=0] shutting down
(kafka.server.KafkaServer)
[2022-04-07 18:19:33,008] INFO [KafkaServer id=0] Starting controlled shutdown
(kafka.server.KafkaServer)
[2022-04-07 18:19:33,016] INFO [KafkaServer id=0] Controlled shutdown request
returned successfully after 6ms (kafka.server.KafkaServer)
....
[2022-04-07 18:19:33,227] INFO Broker and topic stats closed
(kafka.server.BrokerTopicStats)
[2022-04-07 18:19:33,227] INFO App info kafka.server for 0 unregistered
(org.apache.kafka.common.utils.AppInfoParser)
[2022-04-07 18:19:33,227] INFO [KafkaServer id=0] shut down completed
(kafka.server.KafkaServer)
// new added log
[2022-04-07 18:19:33,227] INFO [KafkaServer id=0] the broker is shut down due
to: signal SIGINT{code}
> Log broker shutdown reason at the end of log output
> ---------------------------------------------------
>
> Key: KAFKA-13804
> URL: https://issues.apache.org/jira/browse/KAFKA-13804
> Project: Kafka
> Issue Type: Improvement
> Components: core
> Reporter: Luke Chen
> Assignee: Luke Chen
> Priority: Major
>
> Currently, when the broker hit exception, or SIGINT signal received, we will
> first log the exception (or the reason we started to shutdown) ex:
> _INFO Terminating process due to signal SIGINT_
>
> And then go through the shutdown process, to close all the resources (i.e.
> threads, sockets, metrics...), and in the end, output:
> _[KafkaServer id=0] shut down completed (kafka.server.KafkaServer)_
>
> Sometimes, during closing the resources, we will also hit some exceptions, or
> waiting for some other dependencies close, and output many more logs. When we
> tried to find out why the broker is shutdown from the log, we have to go
> through many lines of logs from the tail, and skip some unrelated exceptions,
> and then identify the specific reason why the broker shutting down.
> We could improve it by adding the reason why the broker is shutting down at
> the end of log, something like this:
>
> {code:java}
> [2022-04-07 18:19:33,005] INFO Terminating process due to signal SIGINT
> (org.apache.kafka.common.utils.LoggingSignalHandler)
> [2022-04-07 18:19:33,007] INFO [KafkaServer id=0] shutting down
> (kafka.server.KafkaServer)
> [2022-04-07 18:19:33,008] INFO [KafkaServer id=0] Starting controlled
> shutdown (kafka.server.KafkaServer)
> [2022-04-07 18:19:33,016] INFO [KafkaServer id=0] Controlled shutdown request
> returned successfully after 6ms (kafka.server.KafkaServer)
> ....
> [2022-04-07 18:19:33,227] INFO Broker and topic stats closed
> (kafka.server.BrokerTopicStats)
> [2022-04-07 18:19:33,227] INFO App info kafka.server for 0 unregistered
> (org.apache.kafka.common.utils.AppInfoParser)
> [2022-04-07 18:19:33,227] INFO [KafkaServer id=0] shut down completed
> (kafka.server.KafkaServer)
> // new added log
> [2022-04-07 18:19:33,227] INFO [KafkaServer id=0] the broker is shut down due
> to: signal SIGINT{code}
>
>
>
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
