~rogpeppe/juju-core/438-local-instance-Addresses : changes

~rogpeppe/juju-core/438-local-instance-Addresses » Changes to store/charmload from revision 1942

From Revision 1922 to 1903

Rev	Summary	Authors	Tags	Date
1922	[r=axwalk] environs/sshstorage: stream stdin [r=axwalk] environs/sshstorage: stream stdin Change sshstorage's Put to stream to the ssh command's stdin, rather than reading into memory and then writing. Also, drive-by: use `mktemp --tmpdir` rather than exporting $TMPDIR. https://codereview.appspot.com/14197043/	Andrew Wilkins		10 years ago
1921	[r=axwalk] Wire up authenticating httpstorage [r=axwalk] Wire up authenticating httpstorage There are two things happening here: - httpstorage authentication changes - wiring up to null provider/localstorage worker. The httpstorage authentication mechanism has changed to use an auth key (randomly generated as part of the boilerplate), rather than client certificates. This gives us more options for using the storage, since now you don't need the client to share the CA key. Also, when serving HTTPS, HTTP is also provided, but only for non-modifying operations. This allows wget et al. to fetch files from the storage without bypassing certificate validation. The HTTP interface disallows modifying operations if there is an HTTPS interface backing it. To get an HTTPS URL, a HEAD request may be performed on the HTTP interface. https://codereview.appspot.com/14011044/	Andrew Wilkins		10 years ago
1920	[r=axwalk] provider/null: define simplestreams tools source [r=axwalk] provider/null: define simplestreams tools source This is somewhat involved, as there's some fixes required (and one drive-by in sshstorage). The change to use GetToolsSources is simple enough, and matches what the local provider does. The Environ override struct in cmd/juju/bootstrap.go had to so that the optional Environ interfaces were still implemented. Overriding like that loses the additional interfaces. To do the above, environs.BootstrapStorager has been changed to have a method that tells the environ to use bootstrap storage, rather than returning a bootstrap storage. The null provider has changed to match, of course, as well as cmd/juju. https://codereview.appspot.com/14011046/	Andrew Wilkins		10 years ago
1919	[r=thumper],[bug=1231724] Make local provider work. [r=thumper],[bug=1231724] Make local provider work. lp:1231724 shows that if a local provider is bootstrapped, the created directory $(JUJU_HOME)/environments is not readable by the user that started the provider as the bootstrap is done by root, and the files and directories created are owned by root with 0600 permission, which means you can't even run juju status on a bootstrapped local provider. https://codereview.appspot.com/14258043/	Tim Penhey		10 years ago
1918	[r=thumper],[bug=1219887] Use a retry strategy to start new ... [r=thumper],[bug=1219887] Use a retry strategy to start new services. On slower disks, upstart doesn't necessarily notice the file straight away. Use a retry strategy to start the newly written upstart file. The first attempt happens straight away, so no slow down on faster systems. https://codereview.appspot.com/14260043/	Tim Penhey		10 years ago
1917	[r=thumper] Really fix the local provider startup. [r=thumper] Really fix the local provider startup. sudo -E causes upstart on saucy to default to use the user bus. --system was added to start and stop, but status was missed. Race conditions during startup masked this before. I've added another check in start so if the actual start fails, check to see if it is started. When the file is written to disk, upstart may start it before we do, particularly, upstart may start it between when we checked it wasn't running, and when we try to start it. A drive by fix to the livetests was added as version.Current was being used to source information from test metadata, and there weren't saucy images defined. The tests pass wtih the version specified with no other alterations. https://codereview.appspot.com/14243043/	Tim Penhey		10 years ago
1916	[r=thumper] Manually clean up lxc autostart symlink [r=thumper] Manually clean up lxc autostart symlink lxc-destroy used to clean this up, now it doensn't. https://codereview.appspot.com/14182044/	Tim Penhey		10 years ago
1915	[r=rogpeppe] environs: add store arg to NewFromName etc [r=rogpeppe] environs: add store arg to NewFromName etc This enables us to delete the EnsureCACert magic. Old certificates should still be used, as the old config magic will still read them. https://codereview.appspot.com/14136043/	Roger Peppe		10 years ago
1914	[r=dimitern] cmd/jujud; others: Disable setting mongo passwd... [r=dimitern] cmd/jujud; others: Disable setting mongo passwds This disables settings mongo passwords for agents. The unit agent is totally cut off from state now, not only it doesn't connect, but it doesn't have a password do to so. The machine agent is a bit more complicated: only machines with JobHostUnits won't have a mongo password or connection to state. Due to firewaller not using the API yet, JobManageEnviron also has to have state connection (and mongo password) for now. We don't do anything yet to reset existing passwords in mongo, but this will come in the following releases. https://codereview.appspot.com/14036045/ R=fwereade, jameinel	Dimiter Naydenov		10 years ago
1913	[r=rogpeppe] cmd/juju: fix test isolation issue [r=rogpeppe] cmd/juju: fix test isolation issue https://codereview.appspot.com/14200044/	Roger Peppe		10 years ago
1912	[r=fwereade] cmd/juju: skip ActualRun tests [r=fwereade] cmd/juju: skip ActualRun tests ...because they seem to be broken everywhere, yet somehow landed all the same: unblocking us is the first step. https://codereview.appspot.com/14202043/	William Reade		10 years ago
1911	[r=jameinel] provider: move to common, drop StartInstance [r=jameinel] provider: move to common, drop StartInstance We originally planned the move from provider to provider/common to include moving StartInstance to environs, but in the end keeping it around seemed like more trouble than it was worth: it looks like a decent utility func, but is only really relevant to the provisioner -- and makes some pretty provisioner-specific and hard-to-justify assumptions about its broker. It's also used extensively in tests, but we have better ways of doing that nowadays. So, the changes should all be categorizable as follows: * move of everything in provider into provider/common * move of funky StartInstance logic into provisioner * deletion of provider.StartInstance * own file, renames for StartInstance helpers in juju/testing * replacement of provider.StartInstance, across the environy tests, with (usually) juju/testing helpers * drive-by removal and comment of provider-type constants in provider * removal of a live test that wasn't testing anything that needed to be tested live (although technically, I guess, it was a direct test of provider.StartInstance, along with a bunch of other stuff) I'm not sure whether I should be testing the lxc broker a bit harder in the light of these changes, or whether the current state is adequate. And there's definitely something up with the BootstrapState file -- the layering is a bit off (we've got cmd/jujud messing using LoadStateFromURL... and I can see why, but it still feels like the naming's a bit off). Not for today. https://codereview.appspot.com/14156045/	William Reade		10 years ago
1910	[r=rogpeppe] cmd/builddb: delete [r=rogpeppe] cmd/builddb: delete It's been broken for a very long time, and we don't need to build mongo with this anymore. https://codereview.appspot.com/14127043/	Roger Peppe		10 years ago
1909	[r=rogpeppe],[bug=1233278] provider/ec2: skip test [r=rogpeppe],[bug=1233278] provider/ec2: skip test See https://bugs.launchpad.net/juju-core/+bug/1233278 https://codereview.appspot.com/14123045/	Roger Peppe		10 years ago
1908	[r=jameinel] Increment Juju to 1.15.1 [r=jameinel] Increment Juju to 1.15.1 1.15.0 is now in the wild. The dev version is now 1.15.1. https://codereview.appspot.com/14092043/	Curtis Hovey		10 years ago
1907	[r=thumper],[bug=1232936] Explicitly specify the system bus ... [r=thumper],[bug=1232936] Explicitly specify the system bus for upstart. With newer upstarts, calling 'sudo -E' to start the local provider caused juju to not find the correct upstart script to start. https://codereview.appspot.com/14114043/	Tim Penhey		10 years ago
1906	[r=thumper],[bug=1210054] Just call destroy for lxc. [r=thumper],[bug=1210054] Just call destroy for lxc. With the current lxc, it calls stop before destroy. This used to work, now (saucy) it errors. Mock has been updated to mirror golxc behaviour. https://codereview.appspot.com/14114044/	Tim Penhey		10 years ago
1905	[r=jameinel] Isolate the test from S3. [r=jameinel] Isolate the test from S3. The test was actually hitting amazon s3 to look for tools, and since the 1.15 tools were added in the last three days, the test suite has been failing since then. Now we mock out the default tools location.	Tim Penhey		10 years ago
1904	[r=natefinch] Fix a couple typos [r=natefinch] Fix a couple typos	Nate Finch		10 years ago
1903	[r=natefinch] MaaS tag support in constraints [r=natefinch] MaaS tag support in constraints Added a new constraint called tags. In MaaS this corresponds to maas-tags, it is currently unsupported in other providers, though it could be implemented in the future to allow a pass-through for provider-specific features. Tags are simply comma-delimited strings. https://codereview.appspot.com/13802045/	Nate Finch	juju-1.15.0	10 years ago