accumulo-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Mike Walch <mwa...@apache.org>
Subject Re: Updates to Accumulo scripts for 2.0
Date Fri, 07 Oct 2016 14:08:54 GMT
I agree with Christopher in that any code that launches Accumulo in
containers or cgroups should be done by downstream projects (that rely on
the Accumulo binary tarball).  For example, I created a project called
Amara (https://github.com/mikewalch/amara) that launches Accumulo in
Marathon/Mesos. While these projects should live in their own repo, it may
make sense to move them to Apache as Accumulo contrib projects if they
become heavily used by the community.

On Thu, Oct 6, 2016 at 6:57 PM Christopher <ctubbsii@apache.org> wrote:

I would strongly prefer we keep our scripts simple, to allow users to
easily wrap them in cgroups, containers, systemd units, etc. Besides,
cgroups don't even exist on all supported platforms. We should not be
baking in cgroups management into the scripts directly. The easier we make
Accumulo scripts, the more users will be able to build on them easily.

On Thu, Oct 6, 2016 at 4:36 PM Jeff Kubina <jeff.kubina@gmail.com> wrote:

> Can we add an option to launch the tservers in specified set of cgroups,
or
> too late?
>
> --
> Jeff Kubina
> 410-988-4436 <(410)%20988-4436> <(410)%20988-4436>
>
>
> On Thu, Oct 6, 2016 at 3:05 PM, Mike Miller <michaelpmiller@gmail.com>
> wrote:
>
> > +1
> > I like the simplicity of the bin directory and the clarity of only
> > having the 4 main scripts. As someone still new to the project, I am a
> > fan of clean up and simplification by any means possible. Granted, I
> > don't have to maintain production clusters or distributions like CDH
> > so I'd imagine others might not share the same sentiment.
> >
> > On Thu, Oct 6, 2016 at 2:17 PM, Mike Walch <mwalch@apache.org> wrote:
> > > Hi Everyone,
> > >
> > > I started work on on simplifying the Accumulo scripts and
configuration
> > > files for the 2.0 release. Given that this change could impact many, I
> > > wanted to solicit feedback early in the process. Below is the JIRA
> issue
> > > for this work and a pull request with some initial code:
> > >
> > > https://issues.apache.org/jira/browse/ACCUMULO-4490
> > > https://github.com/apache/accumulo/pull/160
> > >
> > > While the code in the pull request works, it's not ready to be merged.
> I
> > > plan to do more work on it. If you don't have a lot of time to review
> the
> > > PR, I would read the updated INSTALL.md instructions and take a look
at
> > the
> > > new scripts in 'bin' using links below:
> > >
> > > https://github.com/mikewalch/accumulo/blob/accumulo-4490/INSTALL.md
> > > https://github.com/mikewalch/accumulo/tree/accumulo-4490/assemble/bin
> > >
> > > Many of the old scripts in 'bin' were moved a new 'scripts' directory
> > but I
> > > plan on simplifying or even removing them. In the future, I think
> scripts
> > > in 'bin' should follow semantic versioning and those in 'scripts'
> should
> > be
> > > considered implementation and subject to change. Let me know what you
> > > think.
> > >
> > > -Mike
> >
>

Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message