[
https://issues.apache.org/jira/browse/FLINK-15507?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17119283#comment-17119283
]
Zakelly Lan commented on FLINK-15507:
-------------------------------------
Thanks, [~liyu]. It's a good point for discussion.
Firstly, I think we should also enable the incremental checkpoint by default.
Thus, with default settings, user could leverage the benefit from local
recovery with incremental checkpoint in RocksDB backend, and only a little
additional disk space is consumed. On the other hand, since users choose to use
the RocksDB backend with large state, they should make sure there is enough
local disk space for it. With small disk and large state, RocksDB backend may
not work properly. The task-local full checkpoints aggravate the lack of disk
space, but not the root cause for it. As far as I understand, Flink only keep
the task-local state of the latest checkpoint, so the total disk usage is under
control.
Secondly, I think it's better not to divide the uniform local recovery
configuration into several parts, or to implement different default behaviors
for different backends. Because it may bring confusion to users.
What do you think? [~sewen] [~yunta] [~klion26]
> Activate local recovery for RocksDB backends by default
> -------------------------------------------------------
>
> Key: FLINK-15507
> URL: https://issues.apache.org/jira/browse/FLINK-15507
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / State Backends
> Reporter: Stephan Ewen
> Assignee: Zakelly Lan
> Priority: Blocker
> Labels: pull-request-available
> Fix For: 1.11.0
>
>
> For the RocksDB state backend, local recovery has no overhead when
> incremental checkpoints are used.
> It should be activated by default, because it greatly helps with recovery.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
