Hi William,
We already have the following limits on appendEntries:
key: raft.server.log.appender.buffer.element-limit (int, default=0)
key: raft.server.log.appender.buffer.byte-limit
(org.apache.ratis.util.SizeInBytes, default=4MB (=4194304))
Tsz-Wo
On Thu, Jul 14, 2022 at 5:04 PM William Song <[email protected]> wrote:
> Hi all,
>
> In AppendEntry, leader sends all entries between (followerNext,
> leaderNext) to a follower. When the follower is slow and cannot keep up
> with group’s writing rate, the number of entries in AE RPC will accumulate
> by time, leading to HeapOOM (and/or) DirectOOM. Is there anything we can do
> to avoid this behavior? Maybe we can impose a limit on AE RPC size. What do
> you think?
>
> Regards,
> William Song
>
