I'm just operating in the sandbox VM that MapR put out, not a cluster. I was more concerned about general log writing/recovery. I'm wondering if MapR folks could help with this test.
On Wed, May 22, 2013 at 1:40 PM, Eric Newton <eric.new...@gmail.com> wrote: > Can you verify that it works well in the presence of the agitator? That > bit of code is supposed to fence off a tablet server from its own logs > after the master determines that it is supposed to be dead. > > -Eric > > > > On Wed, May 22, 2013 at 1:27 PM, John Vines <vi...@apache.org> wrote: > >> And it works like a charm. Thanks Eric >> >> >> On Tue, May 21, 2013 at 7:01 PM, John Vines <vi...@apache.org> wrote: >> >> > Sigh, I guess I could do that. I was rushing out the door, but I'll try >> > that. >> > >> > Sent from my phone, please pardon the typos and brevity. >> > Or you could configure the MapR log closer. >> > >> > >> > >> > On Tue, May 21, 2013 at 6:36 PM, John Vines <vi...@apache.org> wrote: >> > >> > > So I finally got around to testing MapR (well, finally got it >> running). >> > > Gave RC4 a test run and, after seeing the stack trace, not surprised >> it >> > > broke in HadoopLogCloser.close(). IllegalStateException gest thrown >> from >> > > line 54 due to the filesystem being of type >> com.mapr.fs.MapRFileSystem. >> > At >> > > this point, I think this should be considered an acceptable break and >> > > something we shoudl try to resolve first thing for 1.5.1. Thoughts? >> > > >> > >> > >
