hi folks! FYI, the Hadoop folks have identified that some recent changes to HDFS have been eating all memory on hosts where their tests run. So that might be the culprit in our recent spat of new odd looking surefire failures.
Yetus is working to add in some guard rails to keep Hadoop from doing this in the future: https://issues.apache.org/jira/browse/YETUS-561 Related, I have a temporary measure in place on our precommit runs that grabs a bunch of machine information (cpus, memory, disks, etc). If you look at build artifacts they'll be in a directory named 'machine' under 'patchprocess'.
