juju/juju

Don't use proxy for lxc images from state server

Modified the wget command to include --no-proxy
so it doesn't attempt to go through the proxy
when contacting the state server to download
the image.

Fix for bug: https://bugs.launchpad.net/juju-core/+bug/1478660
Previously reviewed for 1.25: http://reviews.vapour.ws/r/2267/
1.24: http://reviews.vapour.ws/r/2293/)
1.22: http://reviews.vapour.ws/r/2320/

(Review request: http://reviews.vapour.ws/r/2334/)

Fix 1467037 - Remove camel case username envusers

RemoveEnvironmentUser should convert the username
to lower case when creating the ID for the op.

Previously reviewed here: http://reviews.vapour.ws/r/2268/

(Review request: http://reviews.vapour.ws/r/2336/)

RemoveEnvironmentUser should convert the username
to lower case when creating the ID for the op.

Modified the wget command to include --no-proxy
so it doesn't attempt to go through the proxy
when contacting the state server to download
the image.

Conflicts:
	container/lxc/export_test.go
	container/lxc/lxc.go
	container/lxc/lxc_test.go

Conflicts:
	container/lxc/export_test.go

…-1.23

cmd/jujud/agent: fix occasional TestLoginsDuringUpgrade failures

Retry restricted API attempts because there's a small gap between the post-upgrade version being written to the agent's config and the end of "upgrade mode".

Fixes LP #1432652.

(Review request: http://reviews.vapour.ws/r/1865/)

Retry restricted API attempts because there's a small gap between the
post-upgrade version being written to the agent's config and the end
of "upgrade mode".

Fixes LP #1432652.

Merge pull request #2460 from wallyworld/container-harvest-mode

Container provisioner now supports configured harvest mode

Fixes: https://bugs.launchpad.net/juju-core/+bug/1459885

The container provisioner was hard coded to use harvest mode HarvestDestroyed.
This branch copies a bunch of code from the environment provisioner to allow the container provisioner to instead use the mode configured in environment config.

(Review request: http://reviews.vapour.ws/r/1825/)

Container provisioner now supports configured harvest mode

Fixes: https://bugs.launchpad.net/juju-core/+bug/1459885

The container provisioner was hard coded to use harvest mode HarvestDestroyed.
This branch copies a bunch of code from the environment provisioner to allow the container provisioner to instead use the mode configured in environment config.

Merge pull request #2424 from wallyworld/leadership-err-fix

Fix lease txn structure; do not ignore errors; worker exists on error

Various leadership issues are fixes, as highlighted by WIlliam.

1. The lease persister has been refactored to use the (Juju) standard txnrunner.run() mechanism. This takes a buildTxn function which can be used to ensure that all txn ops are conditional.

2. Based on 1, transaction ops are no longer unconditional.

3. Errors persisting leases are not ignored. The callers receive the error and the worker now exists. This allows it to restart and reload its lease cache to ensure the cache is not corrupted.

4. Unit tests enhanced to check worker exit on error etc.

5. Feature tests enhanced to actually confirm that the lease records are persisted/updated when a lease is claimed / released.

(Review request: http://reviews.vapour.ws/r/1787/)
Conflicts:
	featuretests/leadership_test.go

(Review request: http://reviews.vapour.ws/r/1796/)

Fix lease txn structure; do not ignore errors; worker exists on error

Various leadership issues are fixes, as highlighted by WIlliam.

1. The lease persister has been refactored to use the (Juju) standard txnrunner.run() mechanism. This takes a buildTxn function which can be used to ensure that all txn ops are conditional.

2. Based on 1, transaction ops are no longer unconditional.

3. Errors persisting leases are not ignored. The callers receive the error and the worker now exists. This allows it to restart and reload its lease cache to ensure the cache is not corrupted.

4. Unit tests enhanced to check worker exit on error etc.

5. Feature tests enhanced to actually confirm that the lease records are persisted/updated when a lease is claimed / released.

(Review request: http://reviews.vapour.ws/r/1787/)
Conflicts:
	featuretests/leadership_test.go

Forward port of fix for bug 1454697



(Review request: http://reviews.vapour.ws/r/1770/)

Prune completed transactions

This set of changes integrates the new txn pruning functionality in juju/txn into Juju. A new txnpruner worker for the master machine agent is introduced to call MaybePruneTransactions periodically. Transactions will be pruned if the transaction count has doubled since the last prune.

Fixes LP #1453785.


(Review request: http://reviews.vapour.ws/r/1746/)

Added closure of the channel.

Used a channel instead of a boolean for the patched reportClosedAPI
function.  Also changed the report* functions to take an io.Closer
rather than an interface{}.

Conflicts:
	cmd/jujud/unit_test.go

Agents which use a close worker need to close the API they open
if their start functions fail.  This patch fixes the machine
and unit agents and mocks out the api state upgrader to simulate
the failure and test that the close occurs.

Periodically calls PruneTransactions to keep the txns collection in
check.

github.com/juju/txn.Runner now has a MaybePruneTransactions
method. This change exposes that on multiEnvRunner and as a method on
State.

Runner.MaybePruneTransactions

Revert 2 broken commits in the init service code.

(fixes https://bugs.launchpad.net/juju-core/+bug/1457031)

(Review request: http://reviews.vapour.ws/r/1759/)

…loudinit-init-system-detection"

This reverts commit ba9267f, reversing
changes made to 5ed292b.

…ashisms-1.23"

This reverts commit 3bc3244, reversing
changes made to 5096ca1.

…1.23

Drop bashisms from the discovery script.

(fixes https://bugs.launchpad.net/juju-core/+bug/1457011)

(Review request: http://reviews.vapour.ws/r/1737/)

Merge pull request #2319 from perrito666/fix_1.22_1454599

Added control for expected errors in firwaller (FIX 1454599)

Firewaller should not fail when a machine is not yet provisioned, if such is the case we should just log the occurence.

(Review request: http://reviews.vapour.ws/r/1680/)

(Review request: http://reviews.vapour.ws/r/1730/)

Added control for expected errors in firwaller (FIX 1454599)

Firewaller should not fail when a machine is not yet provisioned, if such is the case we should just log the occurence.

(Review request: http://reviews.vapour.ws/r/1680/)

Fixed lp:1442257 - lxc-default-mtu environ setting added

Removed the existing behavior implemented some time ago - the NICs of
LXC containers created by Juju inherit the MTU setting of their host's
primary NIC. Instead, Juju won't do anything specific for the MTU values
of LXC NICs, except when the new "lxc-default-mtu" environment setting
is explicitly specified as a positive integer value. Then it will be
used for all created LXC (but not KVM) instances and all their network
interfaces.

See bug http://pad.lv/1442257 for more info.

A few drive-by fixes to some of the tests done as well.

Live tested on MAAS with/without address-allocation feature flag.

(Review request: http://reviews.vapour.ws/r/1725/)

Merge pull request #2360 from wallyworld/1.20-client-compatible

Add 'anything' to valid certificate server addresses for 1.20 client compatibility.

Fixes: https://bugs.launchpad.net/juju-core/+bug/1454829

Juju 1.20 clients use "anything" as the server name in their TLS config when making secure connections to the Juju state server. The certificate generation from 1.22 onwards needs to account for this.

(Review request: http://reviews.vapour.ws/r/1719/)

(Review request: http://reviews.vapour.ws/r/1721/)

Add 'anything' to valid certificate server addresses for 1.20 client compatibility.

Fixes: https://bugs.launchpad.net/juju-core/+bug/1454829

Juju 1.20 clients use "anything" as the server name in their TLS config when making secure connections to the Juju state server. The certificate generation from 1.22 onwards needs to account for this.

(Review request: http://reviews.vapour.ws/r/1719/)