robinyqiu commented on a change in pull request #12403:
URL: https://github.com/apache/beam/pull/12403#discussion_r472514238
##########
File path:
sdks/java/io/google-cloud-platform/src/main/java/org/apache/beam/sdk/io/gcp/bigquery/BigQueryServicesImpl.java
##########
@@ -867,6 +872,7 @@ public void deleteDataset(String projectId, String
datasetId)
}
try {
sleeper.sleep(nextBackoffMillis);
+ throttlingMilliSeconds.inc(nextBackoffMillis);
Review comment:
> Yes, we should use backoff1. Rate limit errors only reach to this
point after 2 minutes of backoffs by backoff1 silently inside of the future.
I see. That make sense to me.
> Why do you think it's over-calculated?
Because I am not sure how this metrics is used downstream. I vaguely
remember Dataflow autoscaling will use this number divided by the total time
spent on work item to yield a fraction which signals throttling. If the total
time is not accumulating time spent on all threads then we may over-calculate.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
