[
https://issues.apache.org/jira/browse/KAFKA-6361?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16291812#comment-16291812
]
Jun Rao commented on KAFKA-6361:
--------------------------------
[~hachikuji], thanks for the info. The problem in the description can indeed
happen if the first leader change is due to preferred leader election, in which
case, the ISR won't change.
To address this issue, we could send another OffsetForLeaderEpoch with the
previous leader epoch as you suggested. This may require multiple rounds of
OffsetForLeaderEpoch requests. Another way is to change OffsetForLeaderEpoch
request to send a sequence of (leader epoch, start offset) for the epoch
between the follower's HW and LEO. On the leader side, we find the longest
consecutive sequence of leader epoch whose start offset matches the leader's.
We then return the end offset of the last matching leader epoch.
The above approach doesn't fully fix the issue for a compacted topic. When all
messages for a leader epoch are removed, we may lose the leader epoch. Thus,
the leader epochs between the follower and the leader may not perfectly match.
One way to address this issue is to preserve the offset of the first message in
a leader epoch during log cleaning. This probably can be done separately since
it causes problems rarely.
> Fast leader fail over can lead to log divergence between leader and follower
> ----------------------------------------------------------------------------
>
> Key: KAFKA-6361
> URL: https://issues.apache.org/jira/browse/KAFKA-6361
> Project: Kafka
> Issue Type: Bug
> Reporter: Jason Gustafson
> Assignee: Jason Gustafson
> Labels: reliability
>
> We have observed an edge case in the replication failover logic which can
> cause a replica to permanently fall out of sync with the leader or, in the
> worst case, actually have localized divergence between logs. This occurs in
> spite of the improved truncation logic from KIP-101.
> Suppose we have brokers A and B. Initially A is the leader in epoch 1. It
> appends two batches: one in the range (0, 10) and the other in the range (11,
> 20). The first one successfully replicates to B, but the second one does not.
> In other words, the logs on the brokers look like this:
> {code}
> Broker A:
> 0: offsets [0, 10], leader epoch: 1
> 1: offsets [11, 20], leader epoch: 1
> Broker B:
> 0: offsets [0, 10], leader epoch: 1
> {code}
> Broker A then has a zk session expiration and broker B is elected with epoch
> 2. It appends a new batch with offsets (11, n) to its local log. So we now
> have this:
> {code}
> Broker A:
> 0: offsets [0, 10], leader epoch: 1
> 1: offsets [11, 20], leader epoch: 1
> Broker B:
> 0: offsets [0, 10], leader epoch: 1
> 1: offsets: [11, n], leader epoch: 2
> {code}
> Normally we expect broker A to truncate to offset 11 on becoming the
> follower, but before it is able to do so, broker B has its own zk session
> expiration and broker A again becomes leader, now with epoch 3. It then
> appends a new entry in the range (21, 30). The updated logs look like this:
> {code}
> Broker A:
> 0: offsets [0, 10], leader epoch: 1
> 1: offsets [11, 20], leader epoch: 1
> 2: offsets: [21, 30], leader epoch: 3
> Broker B:
> 0: offsets [0, 10], leader epoch: 1
> 1: offsets: [11, n], leader epoch: 2
> {code}
> Now what happens next depends on the last offset of the batch appended in
> epoch 2. On becoming follower, broker B will send an OffsetForLeaderEpoch
> request to broker A with epoch 2. Broker A will respond that epoch 2 ends at
> offset 21. There are three cases:
> 1) n < 20: In this case, broker B will not do any truncation. It will begin
> fetching from offset n, which will ultimately cause an out of order offset
> error because broker A will return the full batch beginning from offset 11
> which broker B will be unable to append.
> 2) n == 20: Again broker B does not truncate. It will fetch from offset 21
> and everything will appear fine though the logs have actually diverged.
> 3) n > 20: Broker B will attempt to truncate to offset 21. Since this is in
> the middle of the batch, it will truncate all the way to offset 10. It can
> begin fetching from offset 11 and everything is fine.
> The case we have actually seen is the first one. The second one would likely
> go unnoticed in practice and everything is fine in the third case. To
> workaround the issue, we deleted the active segment on the replica which
> allowed it to re-replicate consistently from the leader.
> I'm not sure the best solution for this scenario. Maybe if the leader isn't
> aware of an epoch, it should always respond with {{UNDEFINED_EPOCH_OFFSET}}
> instead of using the offset of the next highest epoch. That would cause the
> follower to truncate using its high watermark. Or perhaps instead of doing
> so, it could send another OffsetForLeaderEpoch request at the next previous
> cached epoch and then truncate using that.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
