Yeah, that could be a consequence. With the cache in place, the calls to delete are async. This isn't generally a problem in an actual installation, but in the integration tests, it can take some time to sync up (depending on the load). I ran it 20 or so times teasing these out, but it's never quite enough. PR pending. :)
On Mon, Oct 23, 2017 at 9:58 AM, Justin Leet <justinjl...@gmail.com> wrote: > See: METRON-1274 <https://issues.apache.org/jira/browse/METRON-1274> and > https://travis-ci.org/apache/metron/builds > > An example failure is https://travis-ci.org/apache/metron/builds/290806887 > > This is possibly intermittent and possibly a result of METRON-1241 > <https://issues.apache.org/jira/browse/METRON-1241>, but it hasn't been > dug > into yet. More details are in the ticket. > > Please refrain from merging PRs into master until this is resolved. > > Thanks, > Justin >
