Hi Brock.
I didn't have any problems with that test, but it's setup to be very
tight(1ms) to encounter a possible bug that I initially encountered. It
took me 20 runs now to get the issue to recreate. It's somewhat
sensitive to HDFS, looking at the output logs from the time it did fail
however, it looks like HDFS was just lagging behind a bit. It definitely
isn't a problem that should hold back the release, but it looks like
it's going to be awkward to get a test that properly tests the sink
without a lot of sleeps and longer idle timeouts... And running unit
tests take long enough as is. Still, if I can't come up with another
alternative I'll make a patch along those lines.
On 11/20/2012 05:52 AM, Brock Noland wrote:
OK, good to hear. I don't feel this racy test requires another RC,
e.g. the flume-1.2.0 file channel tests are quite racy, but if we have
another RC, I'll make sure we get that 1730 is included.
On Mon, Nov 19, 2012 at 2:48 PM, Roshan Naik <[email protected]> wrote:
passed on second attempt. definitely a fuzzy test.
-roshan
On Mon, Nov 19, 2012 at 12:34 PM, Roshan Naik <[email protected]>wrote:
Just once. Following steps..
- git pull --rebase
- mvn clean package
Running it again right now. will keep you posted.
-roshan
