[
https://issues.apache.org/jira/browse/HDFS-15894?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17304740#comment-17304740
]
Xiaoqiao He commented on HDFS-15894:
------------------------------------
Thanks [~prasad-acit] for your proposal. It is useful to track RPC processed
time. IMO, add client ip/hostname and path will be more helpful.
{code:java}
+ private void processDelayedRpc(long begin, String methodname) {
+ long delay = System.currentTimeMillis() - begin;
+ if (delay > rpcRespoceThreshold) {
+ LOG.warn(
+ "Delayed rpc response for " + methodname + " took " + delay + "ms"
+ + " (threshold=" + rpcRespoceThreshold + "ms)");
+ }
{code}
Furthermore, I even think we should add queue time and process time per RPC to
audit log and I have done it in my internal version. But for community there
could be some more compatibility issues. Pending for some more suggestions.
> Trace Time-consuming RPC response of certain threshold.
> -------------------------------------------------------
>
> Key: HDFS-15894
> URL: https://issues.apache.org/jira/browse/HDFS-15894
> Project: Hadoop HDFS
> Issue Type: Improvement
> Reporter: Renukaprasad C
> Assignee: Renukaprasad C
> Priority: Major
> Attachments: HDFS-15894.001.patch, HDFS-15894.002.patch,
> HDFS-15894.003.patch
>
>
> Monitor & Trace Time-consuming RPC requests.
> Sometimes RPC Requests gets delayed, which impacts the system performance.
> Currently, there is no track for delayed RPC request.
> We can log such delayed RPC calls which exceeds certain threshold.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
