[
https://issues.apache.org/jira/browse/FLINK-7783?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16214579#comment-16214579
]
ASF GitHub Bot commented on FLINK-7783:
---------------------------------------
Github user StefanRRichter commented on a diff in the pull request:
https://github.com/apache/flink/pull/4879#discussion_r146159867
--- Diff:
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/ZooKeeperCompletedCheckpointStore.java
---
@@ -163,22 +162,50 @@ public void recover() throws Exception {
LOG.info("Found {} checkpoints in ZooKeeper.",
numberOfInitialCheckpoints);
- for (Tuple2<RetrievableStateHandle<CompletedCheckpoint>,
String> checkpointStateHandle : initialCheckpoints) {
+ // Try and read the state handles from storage. We try until we
either successfully read
+ // all of them or when we reach a stable state, i.e. when
successfully read the same set
+ // of checkpoints in two tries.
--- End diff --
Maybe we could enhance to comment to talk about the reasons why this code
is written in a certain way (DFS outage) and maybe also include a word about
incremental checkpoints to help future maintainers and/or our own memories.
Unfortunately this also signals that this code has to consider lots of implicit
side effects :-(
> Don't always remove checkpoints in ZooKeeperCompletedCheckpointStore#recover()
> ------------------------------------------------------------------------------
>
> Key: FLINK-7783
> URL: https://issues.apache.org/jira/browse/FLINK-7783
> Project: Flink
> Issue Type: Sub-task
> Components: State Backends, Checkpointing
> Affects Versions: 1.4.0, 1.3.2
> Reporter: Aljoscha Krettek
> Assignee: Aljoscha Krettek
> Priority: Blocker
> Fix For: 1.4.0, 1.3.3
>
>
> Currently, we always delete checkpoint handles if they (or the data from the
> DFS) cannot be read:
> https://github.com/apache/flink/blob/91a4b276171afb760bfff9ccf30593e648e91dfb/flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/ZooKeeperCompletedCheckpointStore.java#L180
> This can lead to problems in case the DFS is temporarily not available, i.e.
> we could inadvertently
> delete all checkpoints even though they are still valid.
> A user reported this problem on the mailing list:
> https://lists.apache.org/thread.html/9dc9b719cf8449067ad01114fedb75d1beac7b4dff171acdcc24903d@%3Cuser.flink.apache.org%3E
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
