How do outsiders actually get stuff committed?

[Greg Hill <gr...@RACKSPACE.COM>]

I've had mixed results getting patches into Ambari, so I'd like to know how to go about it better.  In some cases, it's been my fault because I didn't submit it the right way, or the tests didn't pass for some reason.  I've reached out for help to get a working test environment to reproduce failures, and not really gotten much help.  Does someone have a working setup that can run the full test suite on a Mac?  Can you document that process please?  The Wiki instructions to use docker didn't work out.  The tests ran for more than a day before I finally just killed them.  I can mvn test against some of the subfolders like ambari-agent, but others die with inscrutable errors.  For example, ambari-client fails with an error message about not finding ambari-groovy-client in the maven repo (uh, it's what you're testing locally, why are you trying to find it in the repo?).  It's probably an environment issue, but the error tells me nothing useful, and when I asked previously about it, nobody replied.

Apart from those issues, some patches just sit there without feedback.

I have a very simple change to ambari-agent that's been sitting for almost 2 months with only 1 "ship it": https://reviews.apache.org/r/42031/

I have another related ambari-server patch that has no feedback at all: https://reviews.apache.org/r/42032/  Admittedly, I need to add tests there, but there was no existing coverage to add to and I don't know Java well enough to start from nothing to add them.

I'm hoping my most recent one can get some more eyeballs, since I addressed all the feedback I got so far:  https://reviews.apache.org/r/44285/ Hopefully it doesn't end up festering as well.

So, yeah, if you have any advice for someone who isn't a committer to effectively get their changes in, please pass it along.  Do I just need to know who to assign the reviews to or something?  Is there a list somewhere for reference?  I got a couple names yesterday on the mailing list, but reviewboard didn't seem to have them in the list, so I couldn't assign them.

Thanks for any advice.

Greg

Re: How do outsiders actually get stuff committed?

[Alejandro Fernandez <af...@hortonworks.com>]

Dev mailing list doesn't allow attachments. I've uploaded my shell script
here
https://drive.google.com/folderview?id=0By8U4hLO7mEBQ2dvNVBHSHJTTkk&usp=sha
ring

Thanks,
Alejandro

On 3/4/16, 1:07 PM, "Alejandro Fernandez" <af...@hortonworks.com>
wrote:

>I agree about including several reviewers that have worked on that same
>area (git blame), maybe ping them if they don't reply within 24 hours. I
>often wait for 2 reviewers to say "Ship It" even though we only require 1.
>It helps to let someone know that they should submit it once you give the
>green light (in case you wait for more than 1 Ship It).
>
>I'm attaching a shell script I use to copy the files modified in my last
>commit to my vagrant VM.
>I run ./copy_to_vagrant.sh
>to print the commands for the resources I have to copy over.
>I can also execute the script directly from my vagrant VM.
>
>Thanks,
>Alejandro
>
>On 3/3/16, 12:44 PM, "Jonathan Hurley" <jh...@hortonworks.com> wrote:
>
>>There are a lot of patches for Ambari; more than most of us can take a
>>look at during a given day. Adding the "Ambari" group is probably not
>>enough because of this; you'll want to add specific committers to your
>>review list. I'd take a look at the reviews which have been submitted
>>recently and add ~5 people to the list. Try to break it up by area of
>>expertise. For example, I wouldn't add people who mainly work in
>>ambari-web to reviews for ambari-agent since they typically won't have
>>general knowledge to provide a worthwhile review.
>>
>>With all of that said, I think that if you're not a committer, you should
>>add a comment to your reviews when they have enough +1's that you need it
>>committed. Otherwise, the reviewers don't really know.
>>
>>Talking about your development environment, no, I don't think there's any
>>good writeup for getting an Ambari development instance working on a Mac.
>>I do use a Mac and I'm able to successfully get Ambari Server running
>>locally with agents installed on Linux VMs. I have a whole bunch of
>>scripts which I use to copy files around for development purposes; it's
>>not really work documenting this since it's very specific to how I work.
>>But I'm happy to answer questions and share tips and tricks.
>>
>>The maven stuff bugs me too; we have modules which depend on other
>>modules in their same parent project. That's weird; In order to work
>>around this, I have to "mvn clean compile package install -DskipTests".
>>For example, ambari-server depends on ambari-views. In the ambari-views
>>project I need to install the bits in my local maven repo (~/.m2)
>>
>>> On Mar 3, 2016, at 3:26 PM, Greg Hill <gr...@RACKSPACE.COM> wrote:
>>> 
>>> I've had mixed results getting patches into Ambari, so I'd like to know
>>>how to go about it better.  In some cases, it's been my fault because I
>>>didn't submit it the right way, or the tests didn't pass for some
>>>reason.  I've reached out for help to get a working test environment to
>>>reproduce failures, and not really gotten much help.  Does someone have
>>>a working setup that can run the full test suite on a Mac?  Can you
>>>document that process please?  The Wiki instructions to use docker
>>>didn't work out.  The tests ran for more than a day before I finally
>>>just killed them.  I can mvn test against some of the subfolders like
>>>ambari-agent, but others die with inscrutable errors.  For example,
>>>ambari-client fails with an error message about not finding
>>>ambari-groovy-client in the maven repo (uh, it's what you're testing
>>>locally, why are you trying to find it in the repo?).  It's probably an
>>>environment issue, but the error tells me nothing useful, and when I
>>>asked previously about it, nobody replied.
>>> 
>>> Apart from those issues, some patches just sit there without feedback.
>>> 
>>> I have a very simple change to ambari-agent that's been sitting for
>>>almost 2 months with only 1 "ship it":
>>>https://reviews.apache.org/r/42031/
>>> 
>>> I have another related ambari-server patch that has no feedback at all:
>>>https://reviews.apache.org/r/42032/  Admittedly, I need to add tests
>>>there, but there was no existing coverage to add to and I don't know
>>>Java well enough to start from nothing to add them.
>>> 
>>> I'm hoping my most recent one can get some more eyeballs, since I
>>>addressed all the feedback I got so far:
>>>https://reviews.apache.org/r/44285/ Hopefully it doesn't end up
>>>festering as well.
>>> 
>>> So, yeah, if you have any advice for someone who isn't a committer to
>>>effectively get their changes in, please pass it along.  Do I just need
>>>to know who to assign the reviews to or something?  Is there a list
>>>somewhere for reference?  I got a couple names yesterday on the mailing
>>>list, but reviewboard didn't seem to have them in the list, so I
>>>couldn't assign them.
>>> 
>>> Thanks for any advice.
>>> 
>>> Greg
>>
>>
>

Re: How do outsiders actually get stuff committed?

[Alejandro Fernandez <af...@hortonworks.com>]

I agree about including several reviewers that have worked on that same
area (git blame), maybe ping them if they don't reply within 24 hours. I
often wait for 2 reviewers to say "Ship It" even though we only require 1.
It helps to let someone know that they should submit it once you give the
green light (in case you wait for more than 1 Ship It).

I'm attaching a shell script I use to copy the files modified in my last
commit to my vagrant VM.
I run ./copy_to_vagrant.sh
to print the commands for the resources I have to copy over.
I can also execute the script directly from my vagrant VM.

Thanks,
Alejandro

On 3/3/16, 12:44 PM, "Jonathan Hurley" <jh...@hortonworks.com> wrote:

>There are a lot of patches for Ambari; more than most of us can take a
>look at during a given day. Adding the "Ambari" group is probably not
>enough because of this; you'll want to add specific committers to your
>review list. I'd take a look at the reviews which have been submitted
>recently and add ~5 people to the list. Try to break it up by area of
>expertise. For example, I wouldn't add people who mainly work in
>ambari-web to reviews for ambari-agent since they typically won't have
>general knowledge to provide a worthwhile review.
>
>With all of that said, I think that if you're not a committer, you should
>add a comment to your reviews when they have enough +1's that you need it
>committed. Otherwise, the reviewers don't really know.
>
>Talking about your development environment, no, I don't think there's any
>good writeup for getting an Ambari development instance working on a Mac.
>I do use a Mac and I'm able to successfully get Ambari Server running
>locally with agents installed on Linux VMs. I have a whole bunch of
>scripts which I use to copy files around for development purposes; it's
>not really work documenting this since it's very specific to how I work.
>But I'm happy to answer questions and share tips and tricks.
>
>The maven stuff bugs me too; we have modules which depend on other
>modules in their same parent project. That's weird; In order to work
>around this, I have to "mvn clean compile package install -DskipTests".
>For example, ambari-server depends on ambari-views. In the ambari-views
>project I need to install the bits in my local maven repo (~/.m2)
>
>> On Mar 3, 2016, at 3:26 PM, Greg Hill <gr...@RACKSPACE.COM> wrote:
>> 
>> I've had mixed results getting patches into Ambari, so I'd like to know
>>how to go about it better.  In some cases, it's been my fault because I
>>didn't submit it the right way, or the tests didn't pass for some
>>reason.  I've reached out for help to get a working test environment to
>>reproduce failures, and not really gotten much help.  Does someone have
>>a working setup that can run the full test suite on a Mac?  Can you
>>document that process please?  The Wiki instructions to use docker
>>didn't work out.  The tests ran for more than a day before I finally
>>just killed them.  I can mvn test against some of the subfolders like
>>ambari-agent, but others die with inscrutable errors.  For example,
>>ambari-client fails with an error message about not finding
>>ambari-groovy-client in the maven repo (uh, it's what you're testing
>>locally, why are you trying to find it in the repo?).  It's probably an
>>environment issue, but the error tells me nothing useful, and when I
>>asked previously about it, nobody replied.
>> 
>> Apart from those issues, some patches just sit there without feedback.
>> 
>> I have a very simple change to ambari-agent that's been sitting for
>>almost 2 months with only 1 "ship it":
>>https://reviews.apache.org/r/42031/
>> 
>> I have another related ambari-server patch that has no feedback at all:
>>https://reviews.apache.org/r/42032/  Admittedly, I need to add tests
>>there, but there was no existing coverage to add to and I don't know
>>Java well enough to start from nothing to add them.
>> 
>> I'm hoping my most recent one can get some more eyeballs, since I
>>addressed all the feedback I got so far:
>>https://reviews.apache.org/r/44285/ Hopefully it doesn't end up
>>festering as well.
>> 
>> So, yeah, if you have any advice for someone who isn't a committer to
>>effectively get their changes in, please pass it along.  Do I just need
>>to know who to assign the reviews to or something?  Is there a list
>>somewhere for reference?  I got a couple names yesterday on the mailing
>>list, but reviewboard didn't seem to have them in the list, so I
>>couldn't assign them.
>> 
>> Thanks for any advice.
>> 
>> Greg
>
>

Re: How do outsiders actually get stuff committed?

[Jonathan Hurley <jh...@hortonworks.com>]

There are a lot of patches for Ambari; more than most of us can take a look at during a given day. Adding the "Ambari" group is probably not enough because of this; you'll want to add specific committers to your review list. I'd take a look at the reviews which have been submitted recently and add ~5 people to the list. Try to break it up by area of expertise. For example, I wouldn't add people who mainly work in ambari-web to reviews for ambari-agent since they typically won't have general knowledge to provide a worthwhile review. 

With all of that said, I think that if you're not a committer, you should add a comment to your reviews when they have enough +1's that you need it committed. Otherwise, the reviewers don't really know.

Talking about your development environment, no, I don't think there's any good writeup for getting an Ambari development instance working on a Mac. I do use a Mac and I'm able to successfully get Ambari Server running locally with agents installed on Linux VMs. I have a whole bunch of scripts which I use to copy files around for development purposes; it's not really work documenting this since it's very specific to how I work. But I'm happy to answer questions and share tips and tricks.

The maven stuff bugs me too; we have modules which depend on other modules in their same parent project. That's weird; In order to work around this, I have to "mvn clean compile package install -DskipTests". For example, ambari-server depends on ambari-views. In the ambari-views project I need to install the bits in my local maven repo (~/.m2)

> On Mar 3, 2016, at 3:26 PM, Greg Hill <gr...@RACKSPACE.COM> wrote:
> 
> I've had mixed results getting patches into Ambari, so I'd like to know how to go about it better.  In some cases, it's been my fault because I didn't submit it the right way, or the tests didn't pass for some reason.  I've reached out for help to get a working test environment to reproduce failures, and not really gotten much help.  Does someone have a working setup that can run the full test suite on a Mac?  Can you document that process please?  The Wiki instructions to use docker didn't work out.  The tests ran for more than a day before I finally just killed them.  I can mvn test against some of the subfolders like ambari-agent, but others die with inscrutable errors.  For example, ambari-client fails with an error message about not finding ambari-groovy-client in the maven repo (uh, it's what you're testing locally, why are you trying to find it in the repo?).  It's probably an environment issue, but the error tells me nothing useful, and when I asked previously about it, nobody replied.
> 
> Apart from those issues, some patches just sit there without feedback.
> 
> I have a very simple change to ambari-agent that's been sitting for almost 2 months with only 1 "ship it": https://reviews.apache.org/r/42031/
> 
> I have another related ambari-server patch that has no feedback at all: https://reviews.apache.org/r/42032/  Admittedly, I need to add tests there, but there was no existing coverage to add to and I don't know Java well enough to start from nothing to add them.
> 
> I'm hoping my most recent one can get some more eyeballs, since I addressed all the feedback I got so far:  https://reviews.apache.org/r/44285/ Hopefully it doesn't end up festering as well.
> 
> So, yeah, if you have any advice for someone who isn't a committer to effectively get their changes in, please pass it along.  Do I just need to know who to assign the reviews to or something?  Is there a list somewhere for reference?  I got a couple names yesterday on the mailing list, but reviewboard didn't seem to have them in the list, so I couldn't assign them.
> 
> Thanks for any advice.
> 
> Greg