Home | About | Sematext search-lucene.com search-hadoop.com
 Search Hadoop and all its subprojects:

Switch to Threaded View
Accumulo, mail # dev - RestartIT.restartMasterRecovery


Copy link to this message
-
Re: RestartIT.restartMasterRecovery
Billie Rinaldi 2013-09-13, 21:57
If we use :0 it will avoid a race condition where you pick a random open
port but the port gets taken before the monitor binds to it. Plus it
couldn't hurt to make sure the port registration in Zookeeper works.

It sounded like you might be interested in patching this issue. Feel free,
if you'd like. Otherwise I'll probably get to it tomorrow. An example of
retrieving the monitor location is in server.util.Info.

Billie
On Sep 13, 2013 5:37 PM, "Michael Berman" <[EMAIL PROTECTED]> wrote:

> Ok, thanks Billie.
>
> Is it actually important for the MAC to use the :0 binding to get random
> ports?  Or is this just a way to exercise that code path?  It seems like
> this particular test isn't the only consumer that would want to know what
> randomly assigned port MAC services are running on, so I think it would be
> good to have them conveniently accessible somewhere (although admittedly
> the old way of modifying the config map that you passed into the MAConfig
> is a pretty obscure way to expose them).
>
>
> On Fri, Sep 13, 2013 at 5:29 PM, Michael Berman <[EMAIL PROTECTED]> wrote:
>
> > Looks like the RestartIT issue could have been introduced with the
> > introduction of a configurable GC in the MAC in Josh's a028e7 last week,
> > and from the commit log it looks like the 1.5.1 and 1.4.5 branches might
> > have the same issue.
> >
> > Seems like the ReadWriteIT error was introduced in Billie's 6e7269 two
> > days ago.
> >
> > Process question: the relevant issues are all marked resolved.  If I want
> > to submit patches, should I reopen the bugs?  Or file new bugs
> specifically
> > about the ITs?
> >
>