Details
Description
Recently on hbase1-10, it is easy to eat all the space of /tmp. I've checked several times, it is always the stderr*deferred file, which is the temp output of our UTs.
I think it means we have some problem for some of our UTs which could generate very big output, but if the tmp is full, it is not easy for us to find out the root cause as we need to delete the file to recover first...
Looking at the code of surefire
It just creates the file with jdk's Files.createTempFile, the location can be controlled by java.io.tmpdir, so let's point it to the target directory under the jenkins workspace to avoid wat all the spaces of /tmp first.
Attachments
Issue Links
- is related to
-
HBASE-26936 project Jenkins nodes need a fail-safe clean up of surefire deferred output
- Open
- links to