Matthias J. Sax created KAFKA-8957:
--------------------------------------
Summary: Improve docs about `min.isr.` and `acks=all`
Key: KAFKA-8957
URL: https://issues.apache.org/jira/browse/KAFKA-8957
Project: Kafka
Issue Type: Improvement
Components: clients, core
Reporter: Matthias J. Sax
The current docs are as follows:
{code:java}
acks=all
This means the leader will wait for the full set of in-sync replicas to
acknowledge the record. This guarantees that the record will not be lost as
long as at least one in-sync replica remains alive. This is the strongest
available guarantee.{code}
{code:java}
min.in.sync.replicas
When a producer sets acks to "all" (or -1), this configuration specifies the
minimum number of replicas that must acknowledge a write for the write to be
considered successful. If this minimum cannot be met, then the producer will
raise an exception (either NotEnoughReplicas or NotEnoughReplicasAfterAppend).
When used together, `min.insync.replicas` and `acks` allow you to enforce
greater durability guarantees. A typical scenario would be to create a topic
with a replication factor of 3, set min.insync.replicas to 2, and produce with
acks of "all". This will ensure that the producer raises an exception if a
majority of replicas do not receive a write.
{code}
The miss leading part seems to be:
{noformat}
the minimum number of replicas that must acknowledge the write
{noformat}
That could be interpreted to mean that the producer request can return
*_before_* all replicas acknowledge the write. However, min.irs is a
configuration that aims to specify how many replicase must be online, to
consider a partition to be available.
The actual behavior is the following (with replication factor = 3 and min.isr =
2)
* If all three replicas are in-sync, brokers only ack to the producer after
all three replicas got the data. (ie, both follows need to ack)
* However, if one replicas lags (is not in-sync any longer), we are also ok to
ack to the producer after the remaining in-sync follower acked.
It's *_not_* the case, that if all three replicase are in-sync, brokers ack to
the producer after one follower acked to the leader.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
