I'm experimenting with converting all of our MapReduce tests over to use
MiniAccumuloCluster.  It's been really challenging to write tests that
correctly test the behavior of our input and output formats (asserts within
a Mapper don't work as one would hope), but I've gotten all of the test
classes to pass individually.  However, when I run any two of them, the
second one fails with the message: FATAL util.Initialize: Zookeeper needs
to be up and running in order to init. Exiting ...

I can replicate the issue by copying the existing MiniAccumuloClusterTest
to a different name (still ending in Test) and running them both.

Any thoughts?

Billie

Can we test the formats with MR Unit instead of the mini cluster? It's been
a while since I used it, so I could be mistaken.

Sent from my phone, please pardon the typos and brevity.
On Jan 10, 2013 6:33 PM, "Billie Rinaldi" <[EMAIL PROTECTED]> wrote:

> I'm experimenting with converting all of our MapReduce tests over to use
> MiniAccumuloCluster.  It's been really challenging to write tests that
> correctly test the behavior of our input and output formats (asserts within
> a Mapper don't work as one would hope), but I've gotten all of the test
> classes to pass individually.  However, when I run any two of them, the
> second one fails with the message: FATAL util.Initialize: Zookeeper needs
> to be up and running in order to init. Exiting ...
>
> I can replicate the issue by copying the existing MiniAccumuloClusterTest
> to a different name (still ending in Test) and running them both.
>
> Any thoughts?
>
> Billie
>

On Thu, Jan 10, 2013 at 6:35 PM, John Vines <[EMAIL PROTECTED]> wrote:

> Can we test the formats with MR Unit instead of the mini cluster? It's been
> a while since I used it, so I could be mistaken.
>

But I like the MiniAccumuloCluster.  Maybe I could merge them all into one
massive test class.

Billie
>
> Sent from my phone, please pardon the typos and brevity.
> On Jan 10, 2013 6:33 PM, "Billie Rinaldi" <[EMAIL PROTECTED]>
> wrote:
>
> > I'm experimenting with converting all of our MapReduce tests over to use
> > MiniAccumuloCluster.  It's been really challenging to write tests that
> > correctly test the behavior of our input and output formats (asserts
> within
> > a Mapper don't work as one would hope), but I've gotten all of the test
> > classes to pass individually.  However, when I run any two of them, the
> > second one fails with the message: FATAL util.Initialize: Zookeeper needs
> > to be up and running in order to init. Exiting ...
> >
> > I can replicate the issue by copying the existing MiniAccumuloClusterTest
> > to a different name (still ending in Test) and running them both.
> >
> > Any thoughts?
> >
> > Billie
> >
>

On Thu, Jan 10, 2013 at 6:33 PM, Billie Rinaldi
<[EMAIL PROTECTED]> wrote:
> I'm experimenting with converting all of our MapReduce tests over to use
> MiniAccumuloCluster.  It's been really challenging to write tests that
> correctly test the behavior of our input and output formats (asserts within
> a Mapper don't work as one would hope), but I've gotten all of the test
> classes to pass individually.  However, when I run any two of them, the
> second one fails with the message: FATAL util.Initialize: Zookeeper needs
> to be up and running in order to init. Exiting ...
>
> I can replicate the issue by copying the existing MiniAccumuloClusterTest
> to a different name (still ending in Test) and running them both.
>
> Any thoughts?

I will look into it.

Are you running two instances of MiniAccumuloClusterTest from the same
java process or from different Java processes?

>
> Billie

On Fri, Jan 11, 2013 at 9:26 AM, Keith Turner <[EMAIL PROTECTED]> wrote:

> On Thu, Jan 10, 2013 at 6:33 PM, Billie Rinaldi
> <[EMAIL PROTECTED]> wrote:
> > I'm experimenting with converting all of our MapReduce tests over to use
> > MiniAccumuloCluster.  It's been really challenging to write tests that
> > correctly test the behavior of our input and output formats (asserts
> within
> > a Mapper don't work as one would hope), but I've gotten all of the test
> > classes to pass individually.  However, when I run any two of them, the
> > second one fails with the message: FATAL util.Initialize: Zookeeper needs
> > to be up and running in order to init. Exiting ...
> >
> > I can replicate the issue by copying the existing MiniAccumuloClusterTest
> > to a different name (still ending in Test) and running them both.
> >
> > Any thoughts?
>
> I will look into it.
>
> Are you running two instances of MiniAccumuloClusterTest from the same
> java process or from different Java processes?
>

I'm running them by executing a "mvn clean package" at the command line.
They also fail if you select both and run as junit test in eclipse, but
it's slightly different -- the test just seems to hang at the end of the
first test though you see the same FATAL message.  At the command line,
it's clear that it fails at the beginning of the second test.

Billie
>
> >
> > Billie
>

I tried running two MiniAccumuloClusters from the same JVM and it
hung.  This is because I was calling Init in process and some
singletons were being reused.  If I spawn a process to call init, it
works fine.  I will check in a fix.

On Fri, Jan 11, 2013 at 10:00 AM, Billie Rinaldi
<[EMAIL PROTECTED]> wrote:
> On Fri, Jan 11, 2013 at 9:26 AM, Keith Turner <[EMAIL PROTECTED]> wrote:
>
>> On Thu, Jan 10, 2013 at 6:33 PM, Billie Rinaldi
>> <[EMAIL PROTECTED]> wrote:
>> > I'm experimenting with converting all of our MapReduce tests over to use
>> > MiniAccumuloCluster.  It's been really challenging to write tests that
>> > correctly test the behavior of our input and output formats (asserts
>> within
>> > a Mapper don't work as one would hope), but I've gotten all of the test
>> > classes to pass individually.  However, when I run any two of them, the
>> > second one fails with the message: FATAL util.Initialize: Zookeeper needs
>> > to be up and running in order to init. Exiting ...
>> >
>> > I can replicate the issue by copying the existing MiniAccumuloClusterTest
>> > to a different name (still ending in Test) and running them both.
>> >
>> > Any thoughts?
>>
>> I will look into it.
>>
>> Are you running two instances of MiniAccumuloClusterTest from the same
>> java process or from different Java processes?
>>
>
> I'm running them by executing a "mvn clean package" at the command line.
> They also fail if you select both and run as junit test in eclipse, but
> it's slightly different -- the test just seems to hang at the end of the
> first test though you see the same FATAL message.  At the command line,
> it's clear that it fails at the beginning of the second test.
>
> Billie
>
>
>>
>> >
>> > Billie
>>