[
https://issues.apache.org/jira/browse/IMPALA-10180?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17722307#comment-17722307
]
Joe McDonnell commented on IMPALA-10180:
----------------------------------------
There is a bit of complexity here, mainly around how the current metric is
defined. Suppose there is a query execution like this:
{noformat}
Exec()
GetResultSetMetadata()
GetOperationStatus() - not ready to fetch yet
GetOperationStatus() - not ready to fetch yet
...
GetOperationStatus() - ready to fetch
Fetch() - first fetch
Fetch()
Fetch()
Fetch() - end of fetch
... time passes ...
Close / finalize query{noformat}
Currently, ClientFetchWaitTimer covers all the times between the calls from
Exec() all the way through close/finalize. So, I'm thinking about how this
should go.
> Add average size of fetch requests in runtime profile
> -----------------------------------------------------
>
> Key: IMPALA-10180
> URL: https://issues.apache.org/jira/browse/IMPALA-10180
> Project: IMPALA
> Issue Type: Improvement
> Components: Clients
> Reporter: Sahil Takiar
> Assignee: Joe McDonnell
> Priority: Major
>
> When queries with a high {{ClientFetchWaitTimer}} it would be useful to know
> the average number of rows requested by the client per fetch request. This
> can help determine if setting a higher fetch size would help improve fetch
> performance where the network RTT between the client and Impala is high.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
