[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15923781#comment-15923781
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann closed the pull request at:

https://github.com/apache/flink/pull/3446


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
> Fix For: 1.3.0, 1.1.5, 1.2.1
>
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15923779#comment-15923779
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann closed the pull request at:

https://github.com/apache/flink/pull/3451


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
> Fix For: 1.3.0, 1.1.5, 1.2.1
>
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15923778#comment-15923778
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann commented on the issue:

https://github.com/apache/flink/pull/3451
  
Thanks for the review @StephanEwen. See comment on #3446 for the test. 
Merging this PR.


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
> Fix For: 1.3.0, 1.1.5, 1.2.1
>
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15923775#comment-15923775
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann closed the pull request at:

https://github.com/apache/flink/pull/3448


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15923769#comment-15923769
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann commented on the issue:

https://github.com/apache/flink/pull/3446
  
Thanks for the review @StephanEwen. Merging this PR.


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15905282#comment-15905282
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann commented on the issue:

https://github.com/apache/flink/pull/3446
  
The mocking is indeed a little bit complex in this case. The problem is 
that you also want to check that a state handle which fails to retrieve its 
state, is properly discarded. This discard call happens as part of a callback 
to a ZooKeeper remove call. Since the curator client uses the builder API to 
construct the ZooKeeper calls, it was necessary to mock all the different build 
stages. I couldn't find a more succinct way to test this behaviour without 
starting a ZooKeeper server.

If you think it hurts the test's maintainability too much, then I can start 
a ZooKeeper server for the test.


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15905258#comment-15905258
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann commented on a diff in the pull request:

https://github.com/apache/flink/pull/3446#discussion_r105419062
  
--- Diff: 
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/ZooKeeperCompletedCheckpointStore.java
 ---
@@ -226,16 +200,43 @@ public CompletedCheckpoint getLatestCheckpoint() 
throws Exception {
return null;
}
else {
-   return 
checkpointStateHandles.getLast().f0.retrieveState();
+   while(!checkpointStateHandles.isEmpty()) {
+   
Tuple2 
checkpointStateHandle = checkpointStateHandles.peekLast();
+
+   try {
+   return 
retrieveCompletedCheckpoint(checkpointStateHandle);
+   } catch (FlinkException e) {
--- End diff --

Technically, I think it was ok, because the `retrieveCompletedCheckpoint` 
method catches all `Exceptions` and wraps them in a `FlinkException`. But it's 
better to not rely on this implementation detail.


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15905251#comment-15905251
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user tillrohrmann commented on a diff in the pull request:

https://github.com/apache/flink/pull/3446#discussion_r105417543
  
--- Diff: 
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/ZooKeeperCompletedCheckpointStore.java
 ---
@@ -226,16 +200,43 @@ public CompletedCheckpoint getLatestCheckpoint() 
throws Exception {
return null;
}
else {
-   return 
checkpointStateHandles.getLast().f0.retrieveState();
+   while(!checkpointStateHandles.isEmpty()) {
+   
Tuple2 
checkpointStateHandle = checkpointStateHandles.peekLast();
+
+   try {
+   return 
retrieveCompletedCheckpoint(checkpointStateHandle);
+   } catch (FlinkException e) {
--- End diff --

Yes, you're right. Will change it.


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15901987#comment-15901987
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user StephanEwen commented on the issue:

https://github.com/apache/flink/pull/3451
  
Fix looks good, but I have a similar comment about the test as for #3446


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15901975#comment-15901975
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

Github user StephanEwen commented on a diff in the pull request:

https://github.com/apache/flink/pull/3446#discussion_r105021165
  
--- Diff: 
flink-runtime/src/main/java/org/apache/flink/runtime/checkpoint/ZooKeeperCompletedCheckpointStore.java
 ---
@@ -226,16 +200,43 @@ public CompletedCheckpoint getLatestCheckpoint() 
throws Exception {
return null;
}
else {
-   return 
checkpointStateHandles.getLast().f0.retrieveState();
+   while(!checkpointStateHandles.isEmpty()) {
+   
Tuple2 
checkpointStateHandle = checkpointStateHandles.peekLast();
+
+   try {
+   return 
retrieveCompletedCheckpoint(checkpointStateHandle);
+   } catch (FlinkException e) {
--- End diff --

I would catch more than `FlinkException` here - after all, we want to fall 
back to earlier checkpoints in any error case, no?


> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15890631#comment-15890631
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

GitHub user tillrohrmann opened a pull request:

https://github.com/apache/flink/pull/3451

[backport-1.1] [FLINK-5940] [checkpoint] Harden 
ZooKeeperCompletedCheckpointStore.recover method

Backpor of #3446 onto `release-1.1` branch.

The ZooKeeperCompletedCheckpointStore only tries to recover the latest 
completed
checkpoint even though it might have read older checkpoint state handles 
from
ZooKeeper. In order to deal with corrupted state handles, this commit 
changes the
behaviour such that the completed checkpoint store adds all read retrievable
state handles from ZooKeeper and upon request of the latest checkpoint it 
will
return the latest completed checkpoint which could be retrieved from the 
state
handles. Broken state handles are removed from the completed checkpoint 
store and
ZooKeeper.


You can merge this pull request into a Git repository by running:

$ git pull https://github.com/tillrohrmann/flink fixCheckpointRecoveryBp1.1

Alternatively you can review and apply these changes as the patch at:

https://github.com/apache/flink/pull/3451.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

This closes #3451






> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15890590#comment-15890590
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

GitHub user tillrohrmann opened a pull request:

https://github.com/apache/flink/pull/3448

[backport-1.2] [FLINK-5940] [checkpoint] Harden 
ZooKeeperCompletedCheckpointStore.recover method

Backport of #3446 onto `release-1.2` branch.

The ZooKeeperCompletedCheckpointStore only tries to recover the latest 
completed
checkpoint even though it might have read older checkpoint state handles 
from
ZooKeeper. In order to deal with corrupted state handles, this commit 
changes the
behaviour such that the completed checkpoint store adds all read retrievable
state handles from ZooKeeper and upon request of the latest checkpoint it 
will
return the latest completed checkpoint which could be retrieved from the 
state
handles. Broken state handles are removed from the completed checkpoint 
store and
ZooKeeper.



You can merge this pull request into a Git repository by running:

$ git pull https://github.com/tillrohrmann/flink fixCheckpointRecoveryBp1.2

Alternatively you can review and apply these changes as the patch at:

https://github.com/apache/flink/pull/3448.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

This closes #3448


commit f73f5d0bb1491cefa59d1e58ebc8d6a24a4cbaa1
Author: Till Rohrmann 
Date:   2017-03-01T13:08:35Z

[FLINK-5940] [checkpoint] Harden ZooKeeperCompletedCheckpointStore.recover 
method

The ZooKeeperCompletedCheckpointStore only tries to recover the latest 
completed
checkpoint even though it might have read older checkpoint state handles 
from
ZooKeeper. In order to deal with corrupted state handles, this commit 
changes the
behaviour such that the completed checkpoint store adds all read retrievable
state handles from ZooKeeper and upon request of the latest checkpoint it 
will
return the latest completed checkpoint which could be retrieved from the 
state
handles. Broken state handles are removed from the completed checkpoint 
store and
ZooKeeper.




> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)

[jira] [Commented] (FLINK-5940) ZooKeeperCompletedCheckpointStore cannot handle broken state handles

[ASF GitHub Bot (JIRA)]

[ 
https://issues.apache.org/jira/browse/FLINK-5940?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel=15890582#comment-15890582
 ] 

ASF GitHub Bot commented on FLINK-5940:
---

GitHub user tillrohrmann opened a pull request:

https://github.com/apache/flink/pull/3446

[FLINK-5940] [checkpoint] Harden ZooKeeperCompletedCheckpointStore.recover 
method

Before ZooKeeperCompletedCheckpointStore only tried to recover the latest 
completed
checkpoint even though it might have read older checkpoint state handles 
from
ZooKeeper. In order to deal with corrupted state handles, this commit 
changes the
behaviour such that the completed checkpoint store adds all read retrievable
state handles from ZooKeeper and upon request of the latest checkpoint it 
will
return the latest completed checkpoint which could be retrieved from the 
state
handles. Broken state handles are removed from the completed checkpoint 
store and
ZooKeeper.


You can merge this pull request into a Git repository by running:

$ git pull https://github.com/tillrohrmann/flink fixCheckpointRecovery

Alternatively you can review and apply these changes as the patch at:

https://github.com/apache/flink/pull/3446.patch

To close this pull request, make a commit to your master/trunk branch
with (at least) the following in the commit message:

This closes #3446


commit e8fe16004562c012f15f2f1efc0ad45157fbb9dd
Author: Till Rohrmann 
Date:   2017-03-01T13:08:35Z

[FLINK-5940] [checkpoint] Harden ZooKeeperCompletedCheckpointStore.recover 
method

The ZooKeeperCompletedCheckpointStore only tries to recover the latest 
completed
checkpoint even though it might have read older checkpoint state handles 
from
ZooKeeper. In order to deal with corrupted state handles, this commit 
changes the
behaviour such that the completed checkpoint store adds all read retrievable
state handles from ZooKeeper and upon request of the latest checkpoint it 
will
return the latest completed checkpoint which could be retrieved from the 
state
handles. Broken state handles are removed from the completed checkpoint 
store and
ZooKeeper.




> ZooKeeperCompletedCheckpointStore cannot handle broken state handles
> 
>
> Key: FLINK-5940
> URL: https://issues.apache.org/jira/browse/FLINK-5940
> Project: Flink
>  Issue Type: Bug
>  Components: State Backends, Checkpointing
>Affects Versions: 1.2.0, 1.1.4, 1.3.0
>Reporter: Till Rohrmann
>Assignee: Till Rohrmann
>
> The {{ZooKeeperCompletedCheckpointStore}} reads a set of 
> {{RetrievableStateHandles}} from ZooKeeper upon recovery. It then tries to 
> retrieve the {{CompletedCheckpoint}} from the latest state handle. If the 
> retrieve operation fails, then the whole recovery of completed checkpoints 
> fails even though the store might have read older state handles from 
> ZooKeeper. 
> I propose to harden the behaviour by removing broken state handles and 
> returning the first successfully retrieved {{CompletedCheckpoint}}.



--
This message was sent by Atlassian JIRA
(v6.3.15#6346)