zentol commented on a change in pull request #6996: [FLINK-10715] Change
reporter log level
URL: https://github.com/apache/flink/pull/6996#discussion_r230336292
##########
File path:
flink-runtime/src/main/java/org/apache/flink/runtime/metrics/MetricRegistryImpl.java
##########
@@ -426,7 +426,12 @@ public void run() {
try {
reporter.report();
} catch (Throwable t) {
- LOG.warn("Error while reporting metrics", t);
+ if (LOG.isDebugEnabled()) {
+ LOG.debug("Error while reporting
metrics {}", t);
+ }
+ else {
+ LOG.warn("Error while reporting
metrics", t.getMessage());
Review comment:
I'd rather be on the safe side and log the stack-trace in case it is not
caused by us. Whether a report failing or not being critical depends on the
context and user.
For example, a reporter failing consistently over a large period of time
should show up in the logs in a manner that allows the user to make make a
conscious decision whether the cluster has to be restarted or not. However in
the worst case you get a `null` error message, which is entirely unhelpful for
making said decision.
If we want to go for temporary band-aids then iterate over all reporters and
catch `ConcurrentModificationExceptions`. Then you don't have these
somewhat-expected intermittent exceptions caused by us anymore.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
