~frankban/juju-core/minimum-units : changes

~frankban/juju-core/minimum-units » Changes from revision 1292

From Revision 1292 to 1273

Rev	Summary	Authors	Date
1292	Checkpoint. Checkpoint.	Francesco Banconi	11 years ago
1291	Check service life. Check service life.	Francesco Banconi	11 years ago
1290	Merged branch master. Merged branch master.	Francesco Banconi	11 years ago
1289	Merged master. Merged master.	Francesco Banconi	11 years ago
1288	Initial prototype. Initial prototype.	Francesco Banconi	11 years ago
1287	[r=rogpeppe] rpc: allow dynamic change of API [r=rogpeppe] rpc: allow dynamic change of API This will allow us to change the available API in response to a login request. (re-proposed for tarmac; old CL at https://codereview.appspot.com/10148044) https://code.launchpad.net/~rogpeppe/juju-core/317-rpc-dynamic-serve/+merge/1...	Roger Peppe	11 years ago
1286	[r=thumper] Add lxc to our cloud-init packages [r=thumper] Add lxc to our cloud-init packages However don't add lxc if the machine itself is an lxc container as the install fails on precise causing the lxc container to fail to start. https://codereview.appspot.com/10235047/	Tim Penhey	11 years ago
1285	[r=thumper] Add a MachineIdFromTag function. [r=thumper] Add a MachineIdFromTag function. There are times (coming in future branches) where you want to be able to reverse the MachineTag result back into a machineId. https://codereview.appspot.com/10274044/	Tim Penhey	11 years ago
1284	[r=thumper] Move environs.Instance into own package. [r=thumper] Move environs.Instance into own package. Instances are the realizations of state Machines, and could be either containers or environ provider instances. https://codereview.appspot.com/10273044/	Tim Penhey	11 years ago
1283	[r=wallyworld] Juju status supports containers [r=wallyworld] Juju status supports containers juju status shows containers nested against their parent machine. The machine id less than function needed to be updated to support container ids. A drive by additional test for adding containers to an injected machine is also included. https://codereview.appspot.com/10252044/	Matthew Scott	11 years ago
1282	[r=jameinel],[bug=1190560] Ignore Emacs auto-save files in l... [r=jameinel],[bug=1190560] Ignore Emacs auto-save files in lbox pre-checks Emacs auto-save files are named '.#...', so if you have a storage.go file open, Emacs will keep a '.#storage.go' file around for auto-saving. However, the way we pass arguments to gofmt means that '.#' will not be properly escaped and gofmt will not be able to open the file, thus stopping an lbox propose. This change removes any filenames beginning with '.#' with a simple grep. https://codereview.appspot.com/10244044/	Danilo Šegan	11 years ago
1281	[r=jameinel] Fix the cmd/juju tests. [r=jameinel] Fix the cmd/juju tests. One of the tests was emitting an error, but the suite wasn't a logging suite, so it was being output to stderr. Also, there was a file called 2 created in the directory. This was due to an echo >2 line in a plugin test script. This was wrong anyway, we don't emit missing description as it fails other tests. This has been removed. https://codereview.appspot.com/10253053/	Tim Penhey	11 years ago
1280	[r=jameinel] cleaner: implemented the cleaner worker [r=jameinel] cleaner: implemented the cleaner worker The cleaner worker listens to events raised by the cleaner watcher. This signals that there are documents marked for removal. So the worker calls state.Cleanup() to remove them. https://codereview.appspot.com/10148045/	Frank Mueller	11 years ago
1279	[r=jameinel] cmd/juju: Make a test deterministic [r=jameinel] cmd/juju: Make a test deterministic TestGatherDescriptionsInParallel was timing dependent, and occasionally failed on my system. Rather than just increasing the timeout arbitrarily, I changed the code to be more 'event' based. Each plugin requires another plugin to have started before that plugin will return. Thus, if we don't start the plugins in parallel, the plugin will deadlock. Because the failure mode is 'deadlock', I wrapped the call in a select timeout. This means that if it does fail, we have a garbage goroutine and process lying around (but it should never fail). On the plus side, this makes it actually run faster, since each subprocess doesn't waste time sleeping just so we can measure it. (from 350ms down to normally around 15ms.) This is the only test that was failing (based on occasional timing hiccups) for me, which allows us to have this in tarmac.	John Arbash Meinel	11 years ago
1278	Update the validMachine regex for containers Update the validMachine regex for containers Yay regex. R=wallyworld, fwereade CC= https://codereview.appspot.com/10215043	Tim Penhey	11 years ago
1277	deploy refactoring deploy refactoring GUI deploy and CLI deploy are different enough that the common statecmd caused more problems than it solved. Testing is noticeably improved. juju.Conn.DeployService and juju.Conn.AddUnits are now probably ready to move to some other place that just requires a state connection (and not an environment as well); juju.Conn.PutCharm needs some love too, and thought devoted to how we're going to put local charms over the API. But, for now, the various bits all happen in the right place (except that the CLI once again downloads store charms and uploads them itself, rather than taking advantage of that functionality on the server side. This can and will be fixed, but not this CL). R=rog, jameinel, mue CC= https://codereview.appspot.com/10166044	William Reade	11 years ago
1276	state: Service ConfigSettings methods state: Service ConfigSettings methods Config, SetConfig, and SetConfigYAML methods have been dropped in favour of ConfigSettings and UpdateConfigSettings, which use sensible types. Clients are expected to parse their own damn data and supply a sensible format. R=mue, rog CC= https://codereview.appspot.com/10083047	William Reade	11 years ago
1275	state: Service.SetConfigYAML state: Service.SetConfigYAML ...now works as intended. Don't get too attached to it though. R=thumper, wallyworld, rog CC= https://codereview.appspot.com/10172045	William Reade	11 years ago
1274	state: Unit.WatchConfigSettings state: Unit.WatchConfigSettings state.Settings is bad, because it allows any client the ability to read/change the associated document without any sort of validation; and sending them down a channel is independently bad because the documents don't contain the right data for the unit context. So, I just made it an EntityWatcher instead. Works just as well, is rather simpler, doesn't expose massive potential bugs. Win! R=thumper, wallyworld, rog CC= https://codereview.appspot.com/10180043	William Reade	11 years ago
1273	Add ParentId() to Machine Add ParentId() to Machine This branch provides the ability to ask a container for its parent machine's id. R=thumper, fwereade CC= https://codereview.appspot.com/10203043	Ian Booth	11 years ago

Older »