[
https://issues.apache.org/jira/browse/KAFKA-1510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14075447#comment-14075447
]
nicu marasoiu edited comment on KAFKA-1510 at 7/26/14 6:48 PM:
---------------------------------------------------------------
Hi,
isAutoCommit argument works exactly the other way around, apparently it is
"false" from the scheduled auto commit and to "true" from
zkConsConn.commitOffsets()?
So the migration of offsets from zk to kafka is to : set dual commit and kafka
storage, restart consumers, wait for kafka to be copied on the offset commits,
and take out dual commit.
So currently kafka is copied with the offsets only when data flows, and for the
purpose of this task, we need to add one or 2 more cases when it is getting the
offset: when shutting down, or perhaps periodically.
So this task applies only when storage==kafka and dualCommit ==true, right?
I would first ask why the write to zookeeper the new offsets, only if the write
to kafka was ok? My assumption is To make sure only one write to zookeeper,
even though the process of writing to kafka may involve retries.
I would write both directions at all time, and perhaps keep 2 checkpoint
structures, one kafka one zookeeper.
I create a patch now with: a forceCommit that will make that all offsets are
commited to both kafka and zookeeper when shutting down in dual commit mode.
The usefulness of committing all offsets not only to kafka but to zookeeper as
well comes at least from one reason: the one I mentioned above, that if kafka
offset write fails completely, zookeeper is never copied on that.
was (Author: nmarasoi):
Hi,
isAutoCommit argument works exactly the other way around, apparently it is
"false" from the scheduled auto commit and to "true" from
zkConsConn.commitOffsets()?
So the migration of offsets from zk to kafka is to : set dual commit and kafka
storage, restart consumers, wait for kafka to be copied on the offset commits,
and take out dual commit.
So currently kafka is copied with the offsets only when data flows, and for the
purpose of this task, we need to add one or 2 more cases when it is getting the
offset: when shutting down, or perhaps periodically.
So this task applies only when storage==kafka and dualCommit ==true, right?
I would first ask why the write to zookeeper the new offsets, only if the write
to kafka was ok?
I would write both directions at all time, and perhaps keep 2 checkpoint
structures, one kafka one zookeeper.
> Force offset commits when migrating consumer offsets from zookeeper to kafka
> ----------------------------------------------------------------------------
>
> Key: KAFKA-1510
> URL: https://issues.apache.org/jira/browse/KAFKA-1510
> Project: Kafka
> Issue Type: Bug
> Affects Versions: 0.8.2
> Reporter: Joel Koshy
> Assignee: nicu marasoiu
> Labels: newbie
> Fix For: 0.8.2
>
>
> When migrating consumer offsets from ZooKeeper to kafka, we have to turn on
> dual-commit (i.e., the consumers will commit offsets to both zookeeper and
> kafka) in addition to setting offsets.storage to kafka. However, when we
> commit offsets we only commit offsets if they have changed (since the last
> commit). For low-volume topics or for topics that receive data in bursts
> offsets may not move for a long period of time. Therefore we may want to
> force the commit (even if offsets have not changed) when migrating (i.e.,
> when dual-commit is enabled) - we can add a minimum interval threshold (say
> force commit after every 10 auto-commits) as well as on rebalance and
> shutdown.
> Also, I think it is safe to switch the default for offsets.storage from
> zookeeper to kafka and set the default to dual-commit (for people who have
> not migrated yet). We have deployed this to the largest consumers at linkedin
> and have not seen any issues so far (except for the migration caveat that
> this jira will resolve).
--
This message was sent by Atlassian JIRA
(v6.2#6252)
