[
https://issues.apache.org/jira/browse/HDFS-12976?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16500964#comment-16500964
]
Chao Sun commented on HDFS-12976:
---------------------------------
bq. 1. if a non-observer node transitions to observer node, seems such nodes
will not be recognized as available for read by a running client because
observerProxies is set only in constructor.
bq. 2. similarly if an observer node transitions to standby, and even then to
active, seems the write path will not follow this properly Maybe add some logic
to recheck NN status and update the known proxy overtime?
Yes [~vagarychen]: in these two cases the internal state of the proxy provider
will *not* be updated. I also mentioned this in our last sync, and potentially
we should have a better solution than the current one (perhaps refresh the
observer list when {{performFailover}}). This can be done in the follow-up JIRA
for the retry logic.
bq. Another minor side comment, in ObserverReadProxyProvider, this comment does
not seem to fit here: //Find out all the observer proxies
Oops that comment is stale. Anyway, I'll upload a new patch without the retry
and configuration soon, and create follow up jiras for those.
> Introduce StandbyReadProxyProvider
> ----------------------------------
>
> Key: HDFS-12976
> URL: https://issues.apache.org/jira/browse/HDFS-12976
> Project: Hadoop HDFS
> Issue Type: Sub-task
> Components: hdfs-client
> Reporter: Konstantin Shvachko
> Assignee: Chao Sun
> Priority: Major
> Attachments: HDFS-12976.WIP.patch
>
>
> {{StandbyReadProxyProvider}} should implement {{FailoverProxyProvider}}
> interface and be able to submit read requests to ANN and SBN(s).
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
