Hey gang,

Is anyone using the BackupNode and/or CheckpointNode?   I'm not aware
of anyone using them, and the only outstanding related jira is
HDFS-2064, which has been open for over a year and doesn't seem to be
making progress.  For the past three years the code has only been
maintained as part of other changes and test failures, which has
frankly been a bit of a development tax.  Back in 2008 in HADOOP-4539
the goal for the BackupNode was to eventually become a StandbyNode, it
doesn't look like anyone is working on that.  Users get confused as to
whether to use the 2NN or the CheckpointNode, and what the point of
the BackupNode is since it's not provide HA or evolved. Unless someone
wants to push on their evolution and actively maintain them I don't
think it makes sense to keep them in the tree.  Thoughts?

Thanks,
Eli

Agreed, it's been a maintenance burden, and no one using it or actively
maintaining it (except for those of us working on other areas of the code
who have to spend time maintaining it).

-Todd

On Mon, Oct 15, 2012 at 12:12 PM, Eli Collins <[EMAIL PROTECTED]> wrote:

> Hey gang,
>
> Is anyone using the BackupNode and/or CheckpointNode?   I'm not aware
> of anyone using them, and the only outstanding related jira is
> HDFS-2064, which has been open for over a year and doesn't seem to be
> making progress.  For the past three years the code has only been
> maintained as part of other changes and test failures, which has
> frankly been a bit of a development tax.  Back in 2008 in HADOOP-4539
> the goal for the BackupNode was to eventually become a StandbyNode, it
> doesn't look like anyone is working on that.  Users get confused as to
> whether to use the 2NN or the CheckpointNode, and what the point of
> the BackupNode is since it's not provide HA or evolved. Unless someone
> wants to push on their evolution and actively maintain them I don't
> think it makes sense to keep them in the tree.  Thoughts?
>
> Thanks,
> Eli
>

--
Todd Lipcon
Software Engineer, Cloudera

I'm also in favor of removing them. I've fielded many questions from users
about their status, how they compare to other daemons that exist, etc. All
I've seen them do is increase confusion among users and be a maintenance
burden on developers.

--
Aaron T. Myers
Software Engineer, Cloudera

On Mon, Oct 15, 2012 at 12:12 PM, Eli Collins <[EMAIL PROTECTED]> wrote:

> Hey gang,
>
> Is anyone using the BackupNode and/or CheckpointNode?   I'm not aware
> of anyone using them, and the only outstanding related jira is
> HDFS-2064, which has been open for over a year and doesn't seem to be
> making progress.  For the past three years the code has only been
> maintained as part of other changes and test failures, which has
> frankly been a bit of a development tax.  Back in 2008 in HADOOP-4539
> the goal for the BackupNode was to eventually become a StandbyNode, it
> doesn't look like anyone is working on that.  Users get confused as to
> whether to use the 2NN or the CheckpointNode, and what the point of
> the BackupNode is since it's not provide HA or evolved. Unless someone
> wants to push on their evolution and actively maintain them I don't
> think it makes sense to keep them in the tree.  Thoughts?
>
> Thanks,
> Eli
>

It's been 10 days or so: Eli, do you want to file a JIRA to remove this
code?

Do we need to call a vote of any kind, or is this treated as a code change
where the normal processes apply?

-Todd

On Tue, Oct 16, 2012 at 4:51 PM, Aaron T. Myers <[EMAIL PROTECTED]> wrote:

> I'm also in favor of removing them. I've fielded many questions from users
> about their status, how they compare to other daemons that exist, etc. All
> I've seen them do is increase confusion among users and be a maintenance
> burden on developers.
>
> --
> Aaron T. Myers
> Software Engineer, Cloudera
>
>
>
> On Mon, Oct 15, 2012 at 12:12 PM, Eli Collins <[EMAIL PROTECTED]> wrote:
>
> > Hey gang,
> >
> > Is anyone using the BackupNode and/or CheckpointNode?   I'm not aware
> > of anyone using them, and the only outstanding related jira is
> > HDFS-2064, which has been open for over a year and doesn't seem to be
> > making progress.  For the past three years the code has only been
> > maintained as part of other changes and test failures, which has
> > frankly been a bit of a development tax.  Back in 2008 in HADOOP-4539
> > the goal for the BackupNode was to eventually become a StandbyNode, it
> > doesn't look like anyone is working on that.  Users get confused as to
> > whether to use the 2NN or the CheckpointNode, and what the point of
> > the BackupNode is since it's not provide HA or evolved. Unless someone
> > wants to push on their evolution and actively maintain them I don't
> > think it makes sense to keep them in the tree.  Thoughts?
> >
> > Thanks,
> > Eli
> >
>

--
Todd Lipcon
Software Engineer, Cloudera

Filed HDFS-4114. I think it's a regular code change like MR-2736.

Thanks,
Eli

On Fri, Oct 26, 2012 at 7:39 AM, Todd Lipcon <[EMAIL PROTECTED]> wrote:
> It's been 10 days or so: Eli, do you want to file a JIRA to remove this
> code?
>
> Do we need to call a vote of any kind, or is this treated as a code change
> where the normal processes apply?
>
> -Todd
>
> On Tue, Oct 16, 2012 at 4:51 PM, Aaron T. Myers <[EMAIL PROTECTED]> wrote:
>
>> I'm also in favor of removing them. I've fielded many questions from users
>> about their status, how they compare to other daemons that exist, etc. All
>> I've seen them do is increase confusion among users and be a maintenance
>> burden on developers.
>>
>> --
>> Aaron T. Myers
>> Software Engineer, Cloudera
>>
>>
>>
>> On Mon, Oct 15, 2012 at 12:12 PM, Eli Collins <[EMAIL PROTECTED]> wrote:
>>
>> > Hey gang,
>> >
>> > Is anyone using the BackupNode and/or CheckpointNode?   I'm not aware
>> > of anyone using them, and the only outstanding related jira is
>> > HDFS-2064, which has been open for over a year and doesn't seem to be
>> > making progress.  For the past three years the code has only been
>> > maintained as part of other changes and test failures, which has
>> > frankly been a bit of a development tax.  Back in 2008 in HADOOP-4539
>> > the goal for the BackupNode was to eventually become a StandbyNode, it
>> > doesn't look like anyone is working on that.  Users get confused as to
>> > whether to use the 2NN or the CheckpointNode, and what the point of
>> > the BackupNode is since it's not provide HA or evolved. Unless someone
>> > wants to push on their evolution and actively maintain them I don't
>> > think it makes sense to keep them in the tree.  Thoughts?
>> >
>> > Thanks,
>> > Eli
>> >
>>
>
>
>
> --
> Todd Lipcon
> Software Engineer, Cloudera