Description
HADOOP-13871 has shown that network problems can kill perf, and that it's v. hard to track down, even if you turn up the logging in hadoop.fs.s3a and com.amazon layers to debug.
we could maybe improve things by printing out the IPAddress of the s3 endpoint, as that could help with the network tracing. Printing from within hadoop shows the one given to S3a, not a different one returned by any load balancer.
Attachments
Issue Links
- relates to
-
HADOOP-13871 ITestS3AInputStreamPerformance.testTimeToOpenAndReadWholeFileBlocks performance awful
- Resolved
-
HADOOP-14551 S3A init hangs if you try to connect while the system is offline
- Resolved
- links to