lp://staging/~fwereade/juju-core/new-uniter-relations-sketch

Branch merges

Propose for merging

No branches dependent on this one.

Related source package recipes

No recipes using this branch. (?)

Create packaging recipe

Related snap packages

Related bugs

Related blueprints

507. By William Reade on 2012-09-17

uniter relation handling that builds, probably doesn't actually work

506. By Roger Peppe on 2012-09-14

environs/ec2: use different metadata version

Despite being advertised in the amazon documentation, 2012-06-01
is not available in us-east at least. Now we've chosen 2011-01-01
as hopefully being available everywhere. Maybe "latest" would
be better, but who knows what guarantees are provided with
that?

R=fwereade
CC=
https://codereview.appspot.com/6494136

505. By Roger Peppe on 2012-09-14

environs/ec2: use a later version for ec2 metadata

This enables public-hostname and local-hostname to work.

R=fwereade
CC=
https://codereview.appspot.com/6498128

504. By William Reade on 2012-09-14

add git package

R=rog
CC=
https://codereview.appspot.com/6495129

503. By Dave Cheney on 2012-09-14

cmd/juju: add get subcommand

R=fwereade, niemeyer
CC=
https://codereview.appspot.com/6494128

502. By William Reade on 2012-09-13

add toolsDir param to RunHook

I feel this is somewhat inelegant, but clear and simple, and the right step
forward at this stage. I still have ideas floating around my mind for a
hook.Runner, which I currently *expect* to replace HookContext.RunHook, but
I don't want to go off into the weeds looking for the perfect solution when
this one is trivial, and right here.

R=niemeyer
CC=
https://codereview.appspot.com/6496120

501. By Roger Peppe on 2012-09-13

cmd/jujuc/server: log hook stdout/stderr output

The output stops being logged when the hook exits.

R=fwereade, niemeyer
CC=
https://codereview.appspot.com/6494132

500. By William Reade on 2012-09-13

implement charm upgrades

R=niemeyer
CC=
https://codereview.appspot.com/6503072

499. By William Reade on 2012-09-13

combine charm and hook states as uniter state

charm.Manager and hook.StateFile have together been replaced by charm.Deployer
and uniter.StateFile.

The bits of charm.Manager that were not already subsumed by charm.Deployer
moved into uniter.StateFile, as did everything in hook.StateFile; the
overriding benefit of doing this is that it allows us to switch atomically
from a deploying state into a queued-hook state at the end of Uniter.deploy.

The addition of StateFile is technically independent, but I feel it's useful
to see the change as a *move* of the local-state-saving code; apart from those
additions/deletions, the primary feature is a rewritten Uniter.deploy; the
state-reading changes in modes.go are largely mechanical, apart from the move
of a chunk of ModeInit into ModeContinue (because we can now determine
everything about what mode we should be in purely from the state file, there
is no longer any reason to split it across two modes; and ModeInit can keep
doing things that really are about initialization).

In short: don't fear the diff :).

R=niemeyer
CC=
https://codereview.appspot.com/6489083

498. By Roger Peppe on 2012-09-12

container: fix jujud arguments.

Unfortunately we can't test this properly until
we have something using container to do a deployment,
which won't happen until the uniter integration
branch (https://codereview.appspot.com/6497109/)
is merged.

We also add a Unit.AgentName method, meaning
that we can use the same logic in various places that
need to calculate an agent name from a unit.

We also change the path where unit data is stored
($vardir/agents/unit-agent-name) so that any other
agent data can have an entry in the agents directory too.

R=fwereade, niemeyer
CC=
https://codereview.appspot.com/6498117