[
https://issues.apache.org/jira/browse/HDFS-13924?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16655641#comment-16655641
]
Chao Sun commented on HDFS-13924:
---------------------------------
Thanks [~xkrogen] for the great comments!
bq. In FSNamesystem#getBlockLocations, I think the new else-if check should be
guarded by haEnabled && haContext != null, like the one on L1960.
Done.
bq. I don't think ObserverFailedException is the best name for the new
exception. Can we have something more like ObserverRetriableException or
ObserverRetryOnActiveException? I think "Failed" is too strong here; it's not
that the observer failed, just that this particular request needs to be retried.
Done. Changed {{ObserverRetryOnActiveException}}. I think
{{ObserverRetriableException}} is still a little confusing since it's not just
retry but also retry directly on active. :)
bq. The test LGTM, but can we also add a new test to
TestObserverReadProxyProvider ensuring that ObserverFailedException is handled
appropriately in the face of multiple observers?
Done.
bq. For ObserverReadProxyProvider#unwrapRemoteException(), I don't think the
type parameter is doing anything useful, and it seems odd to me that a method
with a name like unwrap... is returning a boolean. I think we should do
something more like the examples of RetryPolicies#isWrappedStandbyException()
or RetryPolicies#getWrappedRetriableException().
Thanks for pointing to {{RetryPolicies#isWrappedStandbyException()}}. Given
that this is only used right now in {{ObserverReadProxyProvider}}, I removed
the function.
bq. Can we enhance the new INFO log message to include info about which
observer proxy threw the exception?
Done.
> Handle BlockMissingException when reading from observer
> -------------------------------------------------------
>
> Key: HDFS-13924
> URL: https://issues.apache.org/jira/browse/HDFS-13924
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Reporter: Chao Sun
> Assignee: Chao Sun
> Priority: Major
> Attachments: HDFS-13924-HDFS-12943.000.patch,
> HDFS-13924-HDFS-12943.001.patch, HDFS-13924-HDFS-12943.002.patch
>
>
> Internally we found that reading from ObserverNode may result to
> {{BlockMissingException}}. This may happen when the observer sees a smaller
> number of DNs than active (maybe due to communication issue with those DNs),
> or (we guess) late block reports from some DNs to the observer. This error
> happens in
> [DFSInputStream#chooseDataNode|https://github.com/apache/hadoop/blob/trunk/hadoop-hdfs-project/hadoop-hdfs-client/src/main/java/org/apache/hadoop/hdfs/DFSInputStream.java#L846],
> when no valid DN can be found for the {{LocatedBlock}} got from the NN side.
> One potential solution (although a little hacky) is to ask the
> {{DFSInputStream}} to retry active when this happens. The retry logic already
> present in the code - we just have to dynamically set a flag to ask the
> {{ObserverReadProxyProvider}} try active in this case.
> cc [~shv], [~xkrogen], [~vagarychen], [~zero45] for discussion.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
