[
https://issues.apache.org/jira/browse/HADOOP-9717?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jing Zhao updated HADOOP-9717:
------------------------------
Attachment: HADOOP-9717.001.patch
Thanks for the review Nicholas! Update the patch based on your comments.
bq. Since protobuf does not have byte or int8
(https://developers.google.com/protocol-buffers/docs/proto#scalar), how about
using int for it in the code?
protobuf uses varint to encode int32/sint32 etc
(https://developers.google.com/protocol-buffers/docs/encoding). Thus using byte
for retry count and keep it within the range 0~0xFE should be able to decrease
the overhead. I changed retryCount from int32 to sint32 since retryCount's
default value is set to -1 currently and sint32 will be more efficient for
negative values.
bq. Server.getCallRetryCount() is not used.
Will HDFS-4979 use this method? I think maybe this method can be called along
with Server#getCallId() and Server#getClientId().
> Add retry flag/retry attempt count to the RPC requests
> ------------------------------------------------------
>
> Key: HADOOP-9717
> URL: https://issues.apache.org/jira/browse/HADOOP-9717
> Project: Hadoop Common
> Issue Type: Improvement
> Components: ipc
> Reporter: Suresh Srinivas
> Assignee: Jing Zhao
> Attachments: HADOOP-9717.000.patch, HADOOP-9717.001.patch
>
>
> RetryCache lookup on server side implementation can be optimized if Rpc
> request indicates if the request is being retried. This jira proposes adding
> an optional field to Rpc request that indicates if request is being retried.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
