[rabbitmq-discuss] rabbit disk_mode branch eating up all RAM, including swap, dying

[Matthew Sackman]

On Mon, Oct 05, 2009 at 03:15:00PM -0500, tsuraan wrote:
> In your current workflow, is bug21444 basically the stable
> accumulation of changes from 21368?  It looks like 21444 is mostly
> merges from 21368 and default, but sometimes the commits indicate that
> 21444 gets its own development as well.  21368 has a ton of
> informative commit messages; it looks more active and less stable.  Is
> that accurate?

No, that's not really accurate. It's difficult to say whether a branch
is "stable" or not. Our basic method is to develop on a branch, then the
branch will go through QA which may well result in further changes to
that branch - that may very well create a cluster of small commits over
the code base as bugs are tracked down or refactorings occur. But at
this point, the code is getting more stable and improving in quality.

21368 is the basic new persister - v2. This is being grown on a
different branch into an altered design (v3) which does not use mnesia
for reasons I've outlined before. 21444 is a set of manual controls that
extend 21368. Thus from time to time I merge 21368 into 21444. All long
standing branches get default merged into them from time to time so that
recent fixes which have been merged into default get propogated out to
branches. This also helps to ensure infrastructure remains working
across all branches, and it makes the eventual merge back into default
easier too.

Matthew