Rebased LXD provider off a good commit from master. #3694

Merged

kat-co merged 607 commits into juju:lxd-provider from kat-co:lxd-provider

Nov 9, 2015

Contributor

kat-co commented Nov 9, 2015

(Review request: http://reviews.vapour.ws/r/3099/)

jujubot and others added some commits Oct 27, 2015


      Merge pull request #3603 from cmars/gccgo-nested-type-fix

Stop using nested types.

They break the gccgo 4.9 compiler.

(Review request: http://reviews.vapour.ws/r/3008/)

737df18


      Renaming ServerTag to be ControllerTag.

b5c4876


      Merge pull request #3604 from alesstimec/cmd-juju-commands-export-dep…

…loy-command-01

Export DeployCommand.



(Review request: http://reviews.vapour.ws/r/3009/)

1a7cf61


      Merge pull request #3602 from voidspace/ignore-addresses-master

1509292: ignore machine addresses

Ignore machine addresses should be ignored for containers, which only have machine addresses.

1f823b8


      Merge pull request #3605 from howbazaar/server-tag-to-controller-tag

Renaming ServerTag to be ControllerTag.

This branch renames the ServerTag functions to be ControllerTag.

The main interesting bit is the params structure returned from LoginResposneV1. This still serializes the ControllerTag as 'server-tag' as this shouldn't change.  When we make the next version of the login response, we can change the serialization format.

A follow-up branch will handle the serverUUID values.

(Review request: http://reviews.vapour.ws/r/3011/)

85cedb8


      Fix documentation typos

c7caec5


      Merge pull request #3597 from howbazaar/add-juju-retry

Add two uses for juju/retry

Add the dependency for the new retry package.

(Review request: http://reviews.vapour.ws/r/3001/)

5fceca5


      Merge pull request #3614 from howbazaar/is-controller-administrator

Rename IsSystemAdministrator to IsControllerAdministrator

Very boring rename.

(Review request: http://reviews.vapour.ws/r/3018/)

32fd892


      worker/uniter/...: test RequestRebootNow

cf4f872


      Update cmd and retry packages.

9dc021e


      api: fix UploadTools to use correct HTTP client logic

	Fixes bugs #1499277 and #1261780


      Merge pull request #3619 from rogpeppe/065-fix-tools-client-api

api: fix UploadTools to use correct HTTP client logic

Fixes bugs #1499277 and #1261780

This allows tools upload to use the new macaroon authorization.


(Review request: http://reviews.vapour.ws/r/3023/)

4e778e4


      update to latest version of dependencies.

Note that this includes the update to persistent-cookiejar that fixes if for concurrent access

c350e23


      Merge pull request #3611 from mjs/fix-typos

Fix documentation typos



(Review request: http://reviews.vapour.ws/r/3016/)

dfb747c


      Merge pull request #3622 from rogpeppe/066-update-deps

update to latest version of dependencies.

Note that this includes the update to persistent-cookiejar that fixes if for concurrent access
and simplifies its API.


(Review request: http://reviews.vapour.ws/r/3026/)

cbbfe37


      Merge master to resolve conflict on imports

82a71f0


      Merge pull request #3472 from bz2/cloudconfig_ubuntu_user

Always create the ubuntu user from cloud-config

Juju expects any ubuntu image to have either an ubuntu user already
created or ubuntu as the default in /etc/cloud/cloud.cfg for
cloud-init to pick up and use.

This change adds cloud-config rules to always create an ubuntu user
with sensible defaults instead of the default user configured in the
image. This enables bootstrap on Dreamhost, and as a side effect
cleans up the centos user creation to share the same configuration.

(Review request: http://reviews.vapour.ws/r/2860/)

2da2ea2


      provider/maas: change ifdown/up sequence when rendering /e/n/i

On aarch64 and kernels >= 3.18, rendering a new /etc/network/interfaces
then running ifdown, ifup causes the existing primary interface to still
have its existing IP address and an associated routing table entry. This
also happens on on amd64/wily/4.2 but on arm64 it is more fatal as
networking is non-functional and juju bootstrap fails.

For example:

$ ifconfig -a
eth0 Link encap:Ethernet HWaddr 52:54:00:c3:72:59
          inet addr:10.17.17.103 Bcast:10.17.17.255 Mask:255.255.255.0

juju-br0 Link encap:Ethernet HWaddr 52:54:00:c3:72:59
          inet addr:10.17.17.103 Bcast:10.17.17.255 Mask:255.255.255.0

Quoting from LP#1496972 (comment #33):

  So the question is, why didn't ifdown kill dhclient? Well, I looked at
  the juju code and I see that it generates a new /e/n/interfaces file
  before running ifdown. I've experienced issues with this before. I
  don't believe that ifdown keeps track of how interfaces were brought
  up itself, it instead consults /e/n/interfaces again. Since /e/n/i was
  already updated, ifdown doesn't know how the primary interface was
  brought up, so fails to bring it down cleanly.

This patch also dispenses with flushing the existing route on the old
primary interface as the new ordering of ifdown, re-render /e/n/i, ifup
will ensure that any existing dhclient processes are terminated and
there will be no extant route to flush.

Thanks to Dann Frazier for the initial patch (see bug/comment #33).

Fixes [LP:#1496972](https://bugs.launchpad.net/juju-core/+bug/1496972)

02a85d6


      Merge pull request #3628 from frobware/master-lp1496972

provider/maas: change ifdown/up sequence when rendering /e/n/i

On aarch64 and kernels >= 3.18, rendering a new /etc/network/interfaces
then running ifdown, ifup causes the existing primary interface to still
have its existing IP address and an associated routing table entry. This
also happens on on amd64/wily/4.2 but on arm64 it is more fatal as
networking is non-functional and juju bootstrap fails.

For example:

$ ifconfig -a
eth0 Link encap:Ethernet HWaddr 52:54:00:c3:72:59
          inet addr:10.17.17.103 Bcast:10.17.17.255 Mask:255.255.255.0

juju-br0 Link encap:Ethernet HWaddr 52:54:00:c3:72:59
          inet addr:10.17.17.103 Bcast:10.17.17.255 Mask:255.255.255.0

Quoting from LP#1496972 (comment #33):

  So the question is, why didn't ifdown kill dhclient? Well, I looked at
  the juju code and I see that it generates a new /e/n/interfaces file
  before running ifdown. I've experienced issues with this before. I
  don't believe that ifdown keeps track of how interfaces were brought
  up itself, it instead consults /e/n/interfaces again. Since /e/n/i was
  already updated, ifdown doesn't know how the primary interface was
  brought up, so fails to bring it down cleanly.

This patch also dispenses with flushing the existing route on the old
primary interface as the new ordering of ifdown, re-render /e/n/i, ifup
will ensure that any existing dhclient processes are terminated and
there will be no extant route to flush.

Thanks to Dann Frazier for the initial patch (see bug/comment #33).

Fixes [LP:#1496972](https://bugs.launchpad.net/juju-core/+bug/1496972)

(Review request: http://reviews.vapour.ws/r/3031/)

adfb74d


      Add bug reference

052b54d


      Updated cookiejar with backwards-compatibility fix.

This revision of persistent-cookiejar will safely handle a cookie file
created by a pre-1.26 Juju release.

e8f5ec7


      Merge pull request #3637 from cmars/fixes-lp1511717

Updated cookiejar with backwards-compatibility fix.

This revision of persistent-cookiejar will safely handle a cookie file
created by a pre-1.26 Juju release.

(Review request: http://reviews.vapour.ws/r/3040/)

eaecfa9


      state: prohibit storing a charm with a nil charm.URL.

Passing a nil charm.URL was not supported, as this would have caused a
panic creating a value for DocID.

This PR adds an assertion ensure that curl is always non nil, and returns
an error when this is not true.

997af91


      environs/config: add specific error for no authorized keys

0da1b68


      Increment juju to 1.26-alpha2

1357af7


      Forwardport old restore deprecation

Forwardport 446b48f from 1.25

df7329e


      Merge pull request #3641 from davecheney/state-add-charm-url-check

state: prohibit storing a charm with a nil charm.URL.

Passing a nil charm.URL was not supported, as this would have caused a
panic creating a value for DocID.

This PR adds an assertion ensure that curl is always non nil, and returns
an error when this is not true.

(Review request: http://reviews.vapour.ws/r/3044/)

fc3963e


      Merge pull request #3579 from cmars/charmdir-fortress

Replace worker/charmdir with worker/fortress.



(Review request: http://reviews.vapour.ws/r/2979/)

0e51912


      Merge pull request #3616 from howbazaar/update-deps

Update cmd and retry packages.

Bring in the latest juju/cmd and juju/retry packages.

(Review request: http://reviews.vapour.ws/r/3020/)

43c10c1


      Merge pull request #3647 from sinzui/inc-1.26-alpha2

Increment juju to 1.26-alpha2



(Review request: http://reviews.vapour.ws/r/3050/)

d256998


      Merge pull request #3598 from axw/uniter-reboot-priority-before-kill

worker/uniter: record reboot before killing hook

When running the "juju-reboot" hook tool with --now, we
kill the hook process and then record the reboot priority.
This is racey: completion of the hook process is the
trigger for the completion of the run-hook operation.

Instead, set the reboot priority first, kill the hook
process, and the revert the reboot priority if killing
the hook process failed.

(Review request: http://reviews.vapour.ws/r/3002/)


      state: added NoTail to LogTailerParams

Setting NoTail make the tailer stop as soon as the logs collection has
been read, without consulting the oplog. This is required for adding a
"no tail" feature to the /logs API and for an in-development utility
for dumping logs directly out of the database.

7be219e


      Merge pull request #3651 from mjs/db-logs-notail

state: added NoTail to LogTailerParams

Setting NoTail makes the tailer stop as soon as the logs collection has been read, without consulting the oplog. This is required for adding a "no tail" feature to the /logs API and for an in-development utility for dumping logs directly out of the database.

(Review request: http://reviews.vapour.ws/r/3054/)

912ccd3


      Inactive meter status worker and tests.

Rework ConnectedStatusWorker into a NotifyWatchHandler.

ca204c4


      Merge pull request #3540 from tasdomas/006-meter-status-inactive-worker

meter status inactive worker

The inactive (for lack of a better name) meter status worker is instantiated when the API connection is unavailable. It's purpose is to escalate the meter status to amber and then to red if the connection does not come back. Although its main goal is simple, most of the complexity in this pull request is dealing with the fact that the worker can be killed and restarted by the dependency engine at any point in time.

(Review request: http://reviews.vapour.ws/r/2938/)

89a2f24


      Merge pull request #3644 from rogpeppe/067-config-ErrNoAuthorizedKeys

environs/config: add specific error for no authorized keys

This allows us to tell when there's been some error
other than just no keys being found.


(Review request: http://reviews.vapour.ws/r/3047/)

809e3ed


      Merge pull request #3648 from perrito666/fix_master_1452082

Forwardport old restore deprecation

Forwardport 446b48f from 1.25

(Review request: http://reviews.vapour.ws/r/3051/)

9ac2a25


      Merge pull request #3260 from natefinch/fix-1486553-1.24

store the charm settings in the same transaction that adds the service

This fixes part of https://bugs.launchpad.net/juju-core/+bug/1486553 **i/o timeout errors can cause non-atomic service deploys** - the part about settings not getting set atomically when the service is added, so you can have a service added without the settings set correctly.

Now the setting and service add are done in the same transaction, so if there's a service, it has the right settings, period.

This does not fix the second part of the problem, where the units may not get assigned, that can still happen (so if you deploy a service with non-zero units, you might get a service with zero units).

(Review request: http://reviews.vapour.ws/r/2639/)
Conflicts:
	state/state_test.go

26f6cac


      change addservice to take a struct, factor out placement code, store …

…placement info on service creation

3f55ba0


      new files

291a092


      assign unit worker and surrounding code

6b5295d


      fix a couple bugs

581c7e1


      fix up the unit add ops so that it doesn't assert the service is aliv…

…e during AddService (since it's not yet). Also change AssignUnits to return a list of the results of the assignation.

7e2330f


      couple refactors to get the unitassignment worker working.

refactor the idprefixwatcher to a generic collection watcher and refactor the watcher to be a strings watcher.

b6edeae


      remove testing panics


      convert a not found constraint into a not found service so we maintai…

…n the same API

bbfc8e4


      fix a bunch of tests

066e68b


      fix a test by making sure we run precheckinstance.. also factored out…

… the logic for converting an instance.Placement into the magic string that provicers want

cd518a9


      all tests passing

Moved the code to run the unitassigner worker into a couple base suites, which fixed a ton of tests.

127a031


      add missing files and some missing comments.

94bf087


      remove comments on code commented out during testing

f5e784a


      fix bad merge

bfac1da


      remove code that causes spurious failures

952fe86


      a little cleanup

a25d372


      use attempt instead of custom retry

c0ebe54


      move some args to a config struct

40895fe


      refactor the service related stuff out of the addunitops, so we can r…

…emove the 'during add service' flag.

13d2837


      move assignment removal to inside assign machine"

5eae3fe


      add todos

dc22fc0


      move the code to insert not founds at the API layer rather than state…

… layer

e712c1e

ericsnowcurrently and others added some commits Oct 27, 2015


      Add Certificate.Fingerprint().


      Add Client.RemoveCert().

3529f00


      Add Client.ListCerts().

4e592b2


      Add Client.RemoveCertByFingerprint().

653d67c


      Remove any added certs.

29542ac


      Drop lxdclient.Client.exposeHostAPI().

1c1827b


      Do not return a Config from finishInstanceConfig.

2db9d4f


      Use the base suite in the LXD provider config tests.

9234ed3


      Drop environConfig.SetRemoteFromHost().

844ac7e


      Use clientCfg.AsNonLocal() in the local case.

27fd5a6


      Add a doc comment to clientServerMethods.

aa38b93


      Factor out prepareRemote().

2d5b85f


      Add a missing doc comment.

24fc577


      Split up (and rename) clientServerMethods.

b71f41b


      Patch out AsNonLocal().

bc7b8f5


      Add Certificate.Name.

0fc94eb


      Add Cert.SetDefaults().

b69e04f


      Add a doc comment for remoteLocalName (and drop remoteDefaultName).


      Clean up some comments.

3e82950


      Add Client.isLocal.

4d7dac9


      Add tests for fixAddr().

55947d1


      Clean up the fixAddr tests.

f1694b8


      Clean up fixAddrs().


      Make fixAddr() more strict.

c148079


      Drop support for unix: URLs.

2d72e42


      SetDefaults -> WithDefaults.

11cbee8


      Add missing build constraints.

7549a83


      make the error from bad connection on local into something more under…

…standable

17db1d9


      added profile methods to lxdclient

c1ac2f5


      added default profile for environ

0d21252


      added default profile to instanceSpec

b17dabd


      Set cert defaults when setting remote defaults.


      Use Certificate.Name.

8ee89b0


      Use a juju-specific cert name.

0e6a49c


      Add a TODO.

2f0f2f6


      Fix tests.

65c14a0


      add some tests

fbadf6b


      touch up

00f2898


      Return a value instead of a pointer from NewCertificate().

e49787e


      Add a note about the implicit copy with a value receiver.

38945b2


      Return a value from Cert.SetDefaults().

f352b13


      AsNonLocal -> UsingTCPRemote.

09a50a1


      Return a value from Remote.UsingTCP().

eded4a0


      Propagate lxdclient changes to the provider.

3e32d8f


      code review changes

a242f86


      SetDefaults -> WithDefaults.

d402d32


      lxd: create juju config directory

a424e08


      Minor clean up and doc comment

6a487d1


      rename Config method and cleanup dir parsing

6d59bd8


      removing unneeded updateLXDVars

a668c0c


      rename to ConfigDir

b78eb00


      updating comments

a6cce79


      Add missing build constraints.

7b99463


      Move prepareRemote to near where it is used.

216a026


      review comments

c70b274


      use remoteidforlocal

5b567e5


      mark tests as go1.3

bfa6920


      Merge branch 'lxd-provider-rebased' into lxd-provider

078e1a1

kat-co added a commit that referenced this pull request Nov 9, 2015


      Merge pull request #3694 from kat-co/lxd-provider

Rebased LXD provider off a good commit from master.

e1f226d

juju/juju

Rebased LXD provider off a good commit from master. #3694

kat-co commented Nov 9, 2015

jujubot and others added some commits Oct 27, 2015

ericsnowcurrently and others added some commits Oct 27, 2015

kat-co added a commit that referenced this pull request Nov 9, 2015

kat-co merged commit `e1f226d` into juju:lxd-provider Nov 9, 2015

juju/juju

Join GitHub today

Rebased LXD provider off a good commit from master. #3694

Conversation

kat-co commented Nov 9, 2015

jujubot and others added some commits Oct 27, 2015

ericsnowcurrently and others added some commits Oct 27, 2015

kat-co added a commit that referenced this pull request Nov 9, 2015

kat-co merged commit e1f226d into juju:lxd-provider Nov 9, 2015

kat-co merged commit `e1f226d` into juju:lxd-provider Nov 9, 2015