[
https://issues.apache.org/jira/browse/FLINK-10357?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16660719#comment-16660719
]
Gary Yao commented on FLINK-10357:
----------------------------------
Here are some problems with the tests that I found:
# We rely on checkpoint statistics to decide when to check the number of files.
However, side effects are triggered asynchronously, i.e., it can happen that
statistics are updated before side effects are triggered. The hashes do not
match because some data is not finalized in time.
# We will likely have to exclude additional exceptions from
{{check_logs_for_exceptions}}.
# Testing job uses fixed delay restart with 3 restart attempts. I suspect that
sometimes this is not enough but I cannot yet explain why.
> Streaming File Sink end-to-end test failed with mismatch
> --------------------------------------------------------
>
> Key: FLINK-10357
> URL: https://issues.apache.org/jira/browse/FLINK-10357
> Project: Flink
> Issue Type: Bug
> Components: Tests
> Affects Versions: 1.6.0
> Reporter: Till Rohrmann
> Assignee: Gary Yao
> Priority: Critical
> Labels: test-stability
> Fix For: 1.6.3, 1.7.0
>
> Attachments: flink-logs.tgz, flink-streaming-file-sink-logs.tgz
>
>
> The {{Streaming File Sink end-to-end test}} failed on an Amazon instance with
> the following result:
> {code}
> FAIL File Streaming Sink: Output hash mismatch. Got
> f2000bbc18a889dc8ec4b6f2b47bf9f5, expected 6727342fdd3aae2129e61fc8f433fb6f.
> head hexdump of actual:
> 0000000 0 \n 1 \n 2 \n 3 \n 4 \n 5 \n 6 \n 7 \n
> 0000010 8 \n 9 \n
> 0000014
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
