charmrevisionupdater retries on error #1564

Closed

voidspace wants to merge 166 commits into juju:master from voidspace:1.21-charmrevup

voidspace commented Feb 9, 2015

The charmrevisionupdater now dies (causing a retry) if there is an error fetching charm revisions. This specifically helps with the case where juju is behind a proxy and needs the proxy settings to be in place before it can access the internet.

gabriel-samfira and others added some commits Oct 29, 2014


      backport reboot tool to 1.21

469f701


      Increment juju to 1.21-beta1

478a7bb


      Merge pull request #1036 from sinzui/inc-1.21-beta1

Increment juju to 1.21-beta1

Increment future stable to 1.21-beta1

cdd587a


      Merge pull request #1038 from waigani/fix-upgrade-bug

Fix upgrade bug

c32a89d


      Merge pull request #1025 from anastasiamac/Bug1320543-debuglog-names-…

…vs-tags

Debug-log accepts names and/or tags for machines/units. Bug 1320543.

2639bcb


      Merge pull request #1041 from anastasiamac/debuglog-tags-and-names

Merge pull request #1025 from anastasiamac/Bug1320543-debuglog-names-vs-tags.

Debug-log accepts names and/or tags for machines/units. 

http://pad.lv/1320543

PR 1025 from master.

3809f55


      Merge pull request #1044 from anastasiamac/Bug1320543-debuglog-names-…

…vs-tags

Go vet warning fix.

dae7698


      Merge pull request #1045 from anastasiamac/debuglog-go-vet

Merge pull request #1044 from anastasiamac/Bug1320543-debuglog-names-vs-tags

Go vet warning fix.

9e5dd4d


      provider/ec2: try next AZ on InsufficientInstanceCapacity

InsufficientInstanceCapacity may be returned by RunInstances
if the AZ cannot allocate instances of the specified instance
type. We will now handle this the same as constrained zones,
by retrying the request with the next least-populated zone.

Fixes https://bugs.launchpad.net/juju-core/+bug/1389037

e409730


      provider/ec2: type switch in isZoneConstrainedError

3c25226


      Merge pull request #1057 from axw/lp1389037-ec2-az-InsufficientInstan…

…ceCapacity-1.21

provider/ec2: try next AZ on InsufficientInstanceCapacity (backport)

(backport to 1.21)

InsufficientInstanceCapacity may be returned by RunInstances
if the AZ cannot allocate instances of the specified instance
type. We will now handle this the same as constrained zones,
by retrying the request with the next least-populated zone.

Fixes https://bugs.launchpad.net/juju-core/+bug/1389037

ab13355


      provider/ec2: try next AZ on "no default subnet" error

If the account/region has a default VPC, but one of the
AZs lacks a default subnet, then Juju may get into a
situation where it cannot provision any instances
successfully. We handle the "no default subnet" error
just as we do constrained zones, by trying the next one.

Fixes https://bugs.launchpad.net/juju-core/+bug/1388860

d09c704


      Merge pull request #1058 from axw/lp1388860-ec2-nodefaultsubnet-1.21

provider/ec2: try next AZ on "no default subnet" error (backport)

(backport to 1.21)

If the account/region has a default VPC, but one of the
AZs lacks a default subnet, then Juju may get into a
situation where it cannot provision any instances
successfully. We handle the "no default subnet" error
just as we do constrained zones, by trying the next one.

Fixes https://bugs.launchpad.net/juju-core/+bug/1388860


      Backported MAAS brige fixes for LXC/KVM containers from master

a4f285c


      Merge pull request #1060 from wallyworld/simplestreams-index2

Add support for using index2.json for tools

Fixes: https://bugs.launchpad.net/juju-core/+bug/1389807

With this change, when loading tools metadata, firstly we look for an index2.json file.
If that's not there, we look for index.json

This allows extra metadata to be added for use by 1.21 and later versions of Juju, and 1.20 and ealrier will not be affected.

The metadata generation plugins have been changed to generate index2.json

As a drive by, some old code, no user used since provider storage was dropped, was deleted.

08cbfd9


      Merge pull request #1059 from dimitern/maas-bridge-fixes-backport-1.21

Backported MAAS brige fixes for LXC/KVM containers from master

See more info in the original PR #1046.
No changes, just backported to 1.21

2dfa668


      Merge pull request #1061 from wallyworld/simplestreams-index2-1.21

Merge pull request #1060 from wallyworld/simplestreams-index2

Add support for using index2.json for tools

Fixes: https://bugs.launchpad.net/juju-core/+bug/1389807

With this change, when loading tools metadata, firstly we look for an index2.json file.
If that's not there, we look for index.json

This allows extra metadata to be added for use by 1.21 and later versions of Juju, and 1.20 and ealrier will not be affected.

The metadata generation plugins have been changed to generate index2.json

As a drive by, some old code, no user used since provider storage was dropped, was deleted.

761aec0


      maas: Fixed doc comments of two helpers

e8b0ecd


      Merge pull request #1063 from dimitern/fix-maas-doc-comments

maas: Fixed doc comments of two helpers

In the previous PR #1046 a couple of helpers methods
in the maas provider got their doc comments trimmed
by mistake. This adds the missing lines.


      Merge pull request #1066 from wallyworld/validate-joyent-credentials

Validate creedentials are correct before attempting to bootstrap a joyent environment

Partially fixes: https://bugs.launchpad.net/bugs/1362072

In a similar vein to what was done for ec2 etc, validate joyent credentials prior to bootstrapping.

Tested live.

f41277b


      Merge pull request #1067 from wallyworld/validate-joyent-credentials-…

…1.21

Merge pull request #1066 from wallyworld/validate-joyent-credentials

Validate creedentials are correct before attempting to bootstrap a joyent environment

Partially fixes: https://bugs.launchpad.net/bugs/1362072

In a similar vein to what was done for ec2 etc, validate joyent credentials prior to bootstrapping.

Tested live.

1a09035


      Add the .reviewboardrc file to help with backporting.


      Merge pull request #1068 from wallyworld/generate-metadata-delete-option

Add --clean option to tools metadata generation

Fixes: https://bugs.launchpad.net/bugs/1390250

Add a --clean option to "juju metadata generate-tools"

This will delete any exiting metadata for the specified stream, so that new metadata replaces instead of merging. Metadata for other streams is untouched.

b938b57


      provider/azure: verify credentials during Prepare

Verify credentials while preparing environment configuration
for Azure, and display a helpful error message on failure.

Fixes https://bugs.launchpad.net/juju-core/+bug/1362072

05a198c


      Fix copy/paste

afc646b


      Merge pull request #1072 from wallyworld/generate-metadata-delete-opt…

…ion-1.21

Merge pull request #1068 from wallyworld/generate-metadata-delete-option

Add --clean option to tools metadata generation

Fixes: https://bugs.launchpad.net/bugs/1390250

Add a --clean option to "juju metadata generate-tools"

This will delete any exiting metadata for the specified stream, so that new metadata replaces instead of merging. Metadata for other streams is untouched.

Also, drive by fix for go vet error.

c085bbd


      Merge pull request #1073 from axw/lp1362072-azure-verify-credentials-…

…1.21

provider/azure: verify credentials during Prepare (backport)

Verify credentials while preparing environment configuration
for Azure, and display a helpful error message on failure.

Fixes https://bugs.launchpad.net/juju-core/+bug/1362072

9aad0dc


      Merge pull request #1071 from ericsnowcurrently/1.21-reviewboardrc

Add the .reviewboardrc file to help with backporting.

a5b1aa8


      Merge pull request #1075 from mjs/1382751-prevent-non-subordinate-con…

…tainer-scoped-relations

1382751 prevent non subordinate container scoped relations

This fixes LP #1382751.

Container scoped relations are now only allowed if there's exactly one subordinate service involved. This prevents conditions that should never have been allowed. The Uniter errors and exits when it encounters relations without a subordinate service on one side.

InferEndpoints now ignores container relations unless there's exactly one subordinate service involved.

a5a84db


      Merge pull request #1054 from anastasiamac/mongo-on-multi-socket-node…

…s-Bug1350337

Wrapping mongod for NUMA machines. Bug 1350337

Bug https://bugs.launchpad.net/juju-core/+bug/1350337

dd1b259


      Merge pull request #1079 from anastasiamac/multinode-numa-fix

Merge pull request #1054 from anastasiamac/mongo-on-multi-socket-nodes-Bug

1350337

Wrapping mongod for NUMA machines. Bug 1350337

Bug https://bugs.launchpad.net/juju-core/+bug/1350337

PR 1054 from master.

21ecfa2


      Remove the new backups CLI.

While the rest of the backups implementation is complete, a key
component (restore) has not landed in time for 1.21.  The existing
restore plugin *should* work with the new backups code, but that has
not been tested at all.  Regardless, it doesn't make a lot of sense
to mix the old and new backups in that way.

Consequently we are effectively disabling the new backups capability
in 1.21.  Completely excising the new backups implementation would
introduce too much risk.  Removing just the CLI portion is sufficient
to effectively eliminate exposing the new backups implementation to
users.  The new API facade is still in place and the various packages
are still there in the code, but that is not likely to be an issue.

This code will remain in the master branch (for 1.22 and forward).

cf1c37e


      Only *unregistering* the backups CLI is necessary.

920dd95


      Merge pull request #1082 from ericsnowcurrently/1.21-excise-backups

Remove the new backups CLI.

While the rest of the backups implementation is complete, a key
component (restore) has not landed in time for 1.21.  The existing
restore plugin *should* work with the new backups code, but that has
not been tested at all.  Regardless, it doesn't make a lot of sense
to mix the old and new backups in that way.

Consequently we are effectively disabling the new backups capability
in 1.21.  Completely excising the new backups implementation would
introduce too much risk.  Removing just the CLI portion is sufficient
to effectively eliminate exposing the new backups implementation to
users.  The new API facade is still in place and the various packages
are still there in the code, but that is not likely to be an issue.

This code will remain in the master branch (for 1.22 and forward).

7da83cb


      merge PR 1048

616f126


      fixes to make stuff compile

350830c


      fix merge error

ad502ff


      Merge pull request #1083 from natefinch/merge-1048

Merge 1048

This merges PR #1048, which is a prerequisite for merging PR #1075 which is needed for 1.21.  The merge for 1075 will come after this one.

f0df85e


      Merge pull request #1077 from wallyworld/backport-subordinate-relatio…

…n-fix

Merge pull request #1075 from mjs/1382751-prevent-non-subordinate-contai...

...ner-scoped-relations

1382751 prevent non subordinate container scoped relations

This fixes LP #1382751.

Container scoped relations are now only allowed if there's exactly one subordinate service involved. This prevents conditions that should never have been allowed. The Uniter errors and exits when it encounters relations without a subordinate service on one side.

InferEndpoints now ignores container relations unless there's exactly one subordinate service involved.

f6ae3d0


      Increment juju to 1.21-beta2

288c311


      Merge pull request #1089 from sinzui/inc-1.21-beta2

Increment juju to 1.21-beta2

Increment 1.21 branch so that new revs can be tested.

9444e6b


      Merge pull request #1090 from wallyworld/unit-public-address-ec2

Unit public address ec2

Fixes: https://bugs.launchpad.net/bugs/1308374

The EC2 provider waits for the instance to advertise its DNS name, and when it does, it inserts this as a public address against the machine.
The unit's get public address method looks at the slice of machine addresses recorded and picks the first public one. If the DNS name is available before the first public IP address is recorded, this will be printed as the public address. So you sometimes get an IP address, sometimes a DNS name.

People want to get just IP addresses for unit's public/private addresses. So the EC2 provider recording of DNS name has been removed. This means that the juju status dns-name value will show an ip address, but that's what happens for HP Cloud and some other providers anyway, and we also want to get rid of dns-name.

Also a drive by fix for a go vet issue.

8adc2a7


      Merge pull request #1091 from wallyworld/unit-public-address-ec2-1.21

Merge pull request #1090 from wallyworld/unit-public-address-ec2

Unit public address ec2

Fixes: https://bugs.launchpad.net/bugs/1308374

The EC2 provider waits for the instance to advertise its DNS name, and when it does, it inserts this as a public address against the machine.
The unit's get public address method looks at the slice of machine addresses recorded and picks the first public one. If the DNS name is available before the first public IP address is recorded, this will be printed as the public address. So you sometimes get an IP address, sometimes a DNS name.

People want to get just IP addresses for unit's public/private addresses. So the EC2 provider recording of DNS name has been removed. This means that the juju status dns-name value will show an ip address, but that's what happens for HP Cloud and some other providers anyway, and we also want to get rid of dns-name.

Also a drive by fix for a go vet issue.

8dbf916


      Merge pull request #1095 from voidspace/maas-1381619

Ignore conflict error when releasing maas nodes

When releasing a node that is not in a releasable status, i.e. typically is already released, maas returns an error 409. Instead of dying and refusing to destroy environment we now ignore that error.

Error 409 is only used for a limited set of statuses, so this seems like safe behaviour. See: https://bugs.launchpad.net/maas/+bug/1381619/comments/15

Fixes bug 1381619.

f6389a9


      Merge pull request #1093 from dimitern/lp-1359714-juju-machine-id-for…

…-charms

Fixed lp:1359714: JUJU_MACHINE_ID added to hook context

JUJU_MACHINE_ID is now available to charm hooks. It contains
the unit's assigned machine ID.

8d12512


      Merge pull request #1121 from wallyworld/backport-lp-1359714

Merge pull request #1093 from dimitern/lp-1359714-juju-machine-id-for-charms

Fixed lp:1359714: JUJU_MACHINE_ID added to hook context

JUJU_MACHINE_ID is now available to charm hooks. It contains
the unit's assigned machine ID.

05ce04d


      Merge pull request #1078 from dimitern/lp-1307677

Fixed #1307677: Local provider network-bridge default respects container type

If "network-bridge" is not specified in environments.yaml for
a local environment, but "container" is explicitly given, this
PR changes the local provider to choose the correct default
network bridge name ("lxcbr0" or "virbr0").

Live tested.

24e147d


      Merge pull request #1094 from dimitern/lp-1382709-openstack-instance-…

…shutoff-not-respected

Fixed lp:1382709: openstack provider now returns SHUTOFF and SUSPENDED instances as well

This fixes http://pad.lv/1382709. When an OpenStack instance
managed by Juju is manually stopped or suspended, its state
changes to SHUTOFF or SUSPENDED, respectively. Because the
openstack provider does not return details for instances
with states other than ACTIVE, BUILD, or BUILD(spawning),
the instance updater worker will fail to retrieve the status
of a manually stopped or suspended instance (there is a warning
in the log instead, which is misleading, as it says "unknown
instance id").

With this patch, instance updater will properly refresh the status
of such instances, after the LongPoll interval (15m) since the last
check has passed. Environ.Instances() for openstack will return
also instances with SHUTOFF or SUSPENDED state.

A quick drive-by fix was done for an annoying log spam at DEBUG
level about floating IP assigned to an instance.

034980f


      Merge pull request #1122 from wallyworld/backport-lp-1307677

Merge pull request #1078 from dimitern/lp-1307677

Fixed #1307677: Local provider network-bridge default respects container type

If "network-bridge" is not specified in environments.yaml for
a local environment, but "container" is explicitly given, this
PR changes the local provider to choose the correct default
network bridge name ("lxcbr0" or "virbr0").

Live tested.

9941e3e


      Merge pull request #1123 from wallyworld/backport-lp-1382709

Merge pull request #1094 from dimitern/lp-1382709-openstack-instance-shutoff-not-respected

Fixed lp:1382709: openstack provider now returns SHUTOFF and SUSPENDED instances as well

This fixes http://pad.lv/1382709. When an OpenStack instance
managed by Juju is manually stopped or suspended, its state
changes to SHUTOFF or SUSPENDED, respectively. Because the
openstack provider does not return details for instances
with states other than ACTIVE, BUILD, or BUILD(spawning),
the instance updater worker will fail to retrieve the status
of a manually stopped or suspended instance (there is a warning
in the log instead, which is misleading, as it says "unknown
instance id").

With this patch, instance updater will properly refresh the status
of such instances, after the LongPoll interval (15m) since the last
check has passed. Environ.Instances() for openstack will return
also instances with SHUTOFF or SUSPENDED state.

A quick drive-by fix was done for an annoying log spam at DEBUG
level about floating IP assigned to an instance.

c99af48


      Abort Remove() if swift container name is empty


      Fix test

6625e41


      Merge pull request #1124 from wallyworld/swift-empty-containername

Abort Remove() if swift container name is empty

Fixes: https://bugs.launchpad.net/bugs/1312217

There have been a couple of goes at this fix. This version does not require any lp.net/goose changes.

From the original PR:



Hi,

I have discovered that through a series of unfortunate events involving a failed destroy-environment run (which I believe was due to a slow Swift API response), and a subsequent "working" (ish) destroy-environment run it is possible for Juju to erroneously send a DELETE HTTP request to the swift account API, which in older versions of swift (and keystone) will result in the account DB for that user in swift being marked as deleted (and the only fix for this is to either a) wait for a cron'd script to clean it up, or b) get someone to manually delete the sqlite3 db for that account hash).

Although I've been unable to find the condition in the environment (e.g. writing/loading the jenv, doing the Remove*(...) calls etc.), but I'm 99.9% recurring that this is the cause for the issue I've had with my Swift account getting into a hard-to-recover state

I've added a simple check to provider/openstack/storage.Remove(...) to ensure the containerName isn't empty (which on a call to delete the container itself, the last step in a destroy env cleanup, will result in a DELETE call to the base URL instead, which swift+keystone will gladly oblige in some OpenStack setups).

I couldn't figure out how to test this correctly, due to the way the Storage instance is created/used in the openstack tests, so any pointers on this would be appreciated.

I feel it's more important to catch this nasty edgecase here above finding it's cause in the env handling purely because it is so destructive in the openstack provider compared to others (e.g. you can't remove the base in S3 so easily).


      Merge pull request #1114 from voidspace/1.21

Merge pull request #1095 from voidspace/maas-1381619

Ignore conflict error when releasing maas nodes

When releasing a node that is not in a releasable status, i.e. typically is already released, maas returns an error 409. Instead of dying and refusing to destroy environment we now ignore that error.

Error 409 is only used for a limited set of statuses, so this seems like safe behaviour. See: https://bugs.launchpad.net/maas/+bug/1381619/comments/15

Fixes bug 1381619.

2eea00b


      Improved svc.SetCharm and unit.SetCharmURL, added more tests

b00dded


      Merge pull request #1129 from dimitern/lp-1301996-backport-for-1.21

Improved svc.SetCharm and unit.SetCharmURL, added more tests

This is an attempt to fix issues like http://pad.lv/1301996

Backported #1113 to 1.21, no other changes.

63d49cc


      Merge pull request #1098 from gabriel-samfira/1.21

backport reboot tool to 1.21

As discussed with @fwereade, this is just a backport of:

#1086

to the 1.21 branch.

efd1d2b


      Merge pull request #1132 from wallyworld/lp-1392514

Ensure metadata file gets closed

Fixes: https://bugs.launchpad.net/juju-core/+bug/1392514

5430a37


      Merge pull request #1133 from wallyworld/lp-1392514-1.21

Merge pull request #1132 from wallyworld/lp-1392514

Ensure metadata file gets closed

Fixes: https://bugs.launchpad.net/juju-core/+bug/1392514

c0c63f3


      worker/uniter/context/jujuc: symlink tools to correct destination

Hook tools were being symlinked to the unit-specific tools symlink
when the first unit was deployed. This causes problems when that
unit is destroyed, and another unit of the same service remains;
the symlinks remain broken.

We change EnsureSymlinks to dereference the target destination if
it is a symlink.

Fixes https://bugs.launchpad.net/juju-core/+bug/1391645

f46fdf8

jujubot and others added some commits Dec 4, 2014


      Merge pull request #1266 from wallyworld/azure-services-for-env

Ensure only instances for this environment are used in azure operations

Instead of using the gwacl method ListPrefixedHostedServices() which can return services from other environments, we use out own method to return services for this environment.

Fixes: https://bugs.launchpad.net/juju-core/+bug/1398820

(Review request: http://reviews.vapour.ws/r/580/)

eb5c903


      Merge pull request #1268 from wallyworld/azure-services-for-env-1.21

Merge pull request #1266 from wallyworld/azure-services-for-env

Ensure only instances for this environment are used in azure operations

Instead of using the gwacl method ListPrefixedHostedServices() which can return services from other environments, we use out own method to return services for this environment.

Fixes: https://bugs.launchpad.net/juju-core/+bug/1398820

(Review request: http://reviews.vapour.ws/r/580/)

(Review request: http://reviews.vapour.ws/r/582/)

c95b406


      Merge pull request #1164 from axw/ec2-blockdevicemapping-allephemeral

provider/ec2: allocate all instance stores

When starting an instance in EC2, specify all of the
instance stores in the block device mapping. The instance
started will have as many four attached, according to
the instance type.

getBlockDeviceMappings will be enhanced in the near
future to support attaching additional volumes to
instances at creation time (i.e. for charm storage).

(Review request: https://reviews.vapour.ws/r/475/)

c77ada7


      Merge pull request #1273 from wallyworld/allocate-all-ec2-stores

Merge pull request #1164 from axw/ec2-blockdevicemapping-allephemeral

provider/ec2: allocate all instance stores

When starting an instance in EC2, specify all of the
instance stores in the block device mapping. The instance
started will have as many four attached, according to
the instance type.

getBlockDeviceMappings will be enhanced in the near
future to support attaching additional volumes to
instances at creation time (i.e. for charm storage).

(Review request: https://reviews.vapour.ws/r/475/)

(Review request: http://reviews.vapour.ws/r/587/)

1e65419


      Backport the fix for lp:1401130 to 1.21 - networker is disabled for j…

…oyent environments

b4fbfd9


      Merge pull request #1309 from dimitern/lp-1401130-disable-networker-o…

…n-joyent-1.21

Backport the fix for lp:1401130 to 1.21 - networker is disabled for joyent environments

No changes other than backporting #1308 for 1.21.

(Review request: http://reviews.vapour.ws/r/624/)

9f6c028


      cloudinit: export AddBootTextFile

This will be needed by the Joyent provider.

e0af8bc


      provider/joyent: add a static route for the internal network

Fixes LP #1401130.

a3b1ac1


      Merge pull request #1318 from mjs/1401130-joyent-routing-1.21

provider/joyent: add a static route for the internal network

Backport from master. Fixes LP #1401130.

(Review request: http://reviews.vapour.ws/r/637/)

cf18d3c


      Change the way the ubuntu user is detected

63bedf1


      cmd/juju: don't use JobManageNetworking with servers that don't suppo…

…rt it

This fixes LP #1402826.

eb8315f


      Merge pull request #1321 from wallyworld/lp-1328958

Change the way the ubuntu user is detected

Fixes https://bugs.launchpad.net/ubuntu/+source/juju-core/+bug/1328958

Instead of looking for /home/ubuntu, we grep the passwd file to determine if the ubuntu user exists.

(Review request: http://reviews.vapour.ws/r/640/)

5fe0672


      Merge pull request #1322 from mjs/1402826-jobmanagenetworking-compat-…

…1.21

cmd/juju: don't use JobManageNetworking with servers that don't support it

This fixes LP #1402826.

(Review request: http://reviews.vapour.ws/r/641/)

5d9c8dc


      Fixed lp:1397376 - Backported to 1.21

cf8ab30


      Merge pull request #1324 from dimitern/lp-1397376-backport-to-1.21

Fixed lp:1397376 - Backported to 1.21

Backport of #1297 to fix http://pad.lv/1397376 for 1.21.
No changes other than adapting the code for 1.21, except
for upgrading juju/testing version used to the latest, so
JSONEquals and YAMLEquals checkers can be used.

Live tested on EC2, Canonistack, Local, MAAS.

(Review request: http://reviews.vapour.ws/r/643/)

304802e


      mongo: check all addresses when deciding if machine is master

The recent fix for LP #1397376 meant that on MAAS the replicaset
master address no longer matches the address returned by
SelectPeerAddress - no state server thought it was the master. This
breaks upgrades among other things.

All machine addresses are no compared against the replicaset master
address.

Fixes LP #1403200.

c9d9847


      Merge pull request #1328 from mjs/1403200-maas-upgrades-1.21

mongo: check all addresses when deciding if machine is master

The recent fix for LP #1397376 meant that on MAAS the replicaset master address no longer matches the address returned by SelectPeerAddress - no state server thought it was the master. This breaks upgrades among other things.

All machine addresses are no compared against the replicaset master address.

Fixes LP #1403200.

(Review request: http://reviews.vapour.ws/r/647/)

c515fb4


      cmd/jujud: fixed patch race in upgrade tests (backport to 1.21)

The patch to the attempt strategy used during upgrades was done just
/after/ the upgrade worker had already been started so there was a
race. The worker was occasionally getting the real attempt strategy
instead of the test one, causing long test runs and unexpected numbers
of retries.

The patch is now made before starting the worker.

Fixes LP #1389389.

0b4dfcd


      Merge pull request #1330 from mjs/1389389-TestUpgradeStepsFailure-int…

…ermittent-failure-1.21

cmd/jujud: fixed patch race in upgrade tests (backport to 1.21)

The patch to the attempt strategy used during upgrades was done just /after/ the upgrade worker had already been started so there was a race. The worker was occasionally getting the real attempt strategy instead of the test one, causing long test runs and unexpected numbers of retries.

The patch is now made before starting the worker.

Fixes LP #1389389.

(Review request: http://reviews.vapour.ws/r/649/)

207f076


      Increment juju to 1.21-beta5

ca438e3


      Merge pull request #1332 from sinzui/inc-1.21.beta5

Increment juju to 1.21-beta5

Update juju and win installer 1.21.beta5, we hope to rename to 1.21.0 after testing

(Review request: http://reviews.vapour.ws/r/650/)


      Fixed lp:1394524 - Backported for 1.21 - apt-get retrying works

9b6187a


      Merge pull request #1338 from dimitern/lp-1394524-backport-1.21

Fixed lp:1394524 - Backported for 1.21 - apt-get retrying works

Original PR: #1195

Bug: http://pad.lv/1394524

Since juju/utils changed a lot since 1.21-beta3 was released, the
backport needed to include a few other changes, e.g. make sure
set.NewStrings() is used everywhere before calling set.Add() as it now
panics, as well as a few changes around backups.

Live tested on local, maas, ec2.

(Review request: http://reviews.vapour.ws/r/656/)


      Merge pull request #1397 from wallyworld/maas-node-constraints

Ensure start instance uses constraints for the MAAS provider

When MAAS provider was starting instances, it was neglecting to pass through any constraints. Oops.

(Review request: http://reviews.vapour.ws/r/718/)

48eb3df


      Merge pull request #1400 from wallyworld/maasstartinstance-test-failure

Remove arch from constraint used in test

Remove arch from a constraint used in a test to allow the test to run with fake tools on all platforms.

(Review request: http://reviews.vapour.ws/r/721/)

4d7800b


      Some test fixes due to old code

9d52ac7


      Merge pull request #1416 from wallyworld/backport-constraints-fix

Backport constraints fix

Fixes: https://bugs.launchpad.net/bugs/1408762

Backport fix from 1.22

(Review request: http://reviews.vapour.ws/r/737/)

7c452f7


      Increment juju to 1.21-rc1

79d8328


      Merge pull request #1424 from sinzui/inc-1.21-rc1

Increment juju to 1.21-rc1

We are calling this version rc1 in case we need to release a fix for a regression. We hope to rename this to 1.21.0 after stakeholders bless it.

(Review request: http://reviews.vapour.ws/r/744/)

8abac93


      cmd/jujud: introduce api-post-upgrade worker

This worker runs all other workers that need to be started after
upgrades have finished. The main benefit is that it stops container
setup from running during upgrades, but it also makes the code more
readable and avoids many startWorkerAfterUpgrade calls.

Part of the fix for LP #1403738.

Conflicts:
	cmd/jujud/machine.go

7d3edab


      Merge pull request #1432 from ericsnowcurrently/1.21-backport-fix-140…

…3738

cmd/jujud: introduce api-post-upgrade worker

See https://bugs.launchpad.net/juju-core/1.21/+bug/1411502.

The 1.22 patch (for 1403738) is PR #1343 (merged: 3cecce6).

(Review request: http://reviews.vapour.ws/r/752/)

dd57286


      upgrades: fix steps iteration when current version has a tag

Don't report an upgrade as being necessary when the target version has
a tag but is the same as the "from" version.

Fixes LP #1411502.

2b05f58


      Merge pull request #1433 from mjs/1411502-upgrade-version-tag-handlin…

…g-1.21

upgrades: fix steps iteration when current version has a tag

Don't report an upgrade as being necessary when the target version has a tag but is the same as the "from" version.

Fixes LP #1411502.

(Review request: http://reviews.vapour.ws/r/753/)

32838a9


      Increment juju to 1.21-rc2

5df095f


      Merge pull request #1447 from sinzui/inc-1.21-rc2

Increment juju to 1.21-rc2

We hope to never use this version, We hope to rename this 1.21.0  after a round of testing.

(Review request: http://reviews.vapour.ws/r/768/)

a8cfe6f


      Increment version to 1.21.0 and add fallback vivid support


      Merge pull request #1462 from wallyworld/inc-1.21

Increment version to 1.21.0 and add fallback vivid support

Increments version number to 1.21.0

Also fixes https://bugs.launchpad.net/juju-core/1.21/+bug/1413245
(The vivid line in support series info was accidentally cut out).


(Review request: http://reviews.vapour.ws/r/783/)

6188fdf


      Increment juju to 1.21.1

96caf89


      Merge pull request #1473 from sinzui/inc-1.21.1

Increment juju to 1.21.1

Update future stable to a possible micro release.

(Review request: http://reviews.vapour.ws/r/795/)

1d283c1


      Add yosemite support

8d5579c


      Merge pull request #1488 from wallyworld/backport-yosemite

Add yosemite support

Backport fix to support yosemite on darwin.

(Review request: http://reviews.vapour.ws/r/810/)

f348323


      Increment juju to 1.21.2

dfdcb85


      Merge pull request #1496 from sinzui/inc-1.21.2

Increment juju to 1.21.2

We fixed the only known 1.21 bug. We will make 1.21.1 stable tomorrow, so re increment to 1.21.2

(Review request: http://reviews.vapour.ws/r/818/)

8a74309


      Fixed lp:1416134 - backport for 1.21

bda59b5


      Merge pull request #1510 from dimitern/lp-1416134-local-kvm-network-b…

…ridge-1.21

Fixed lp:1416134 - backport for 1.21

Just a backport of #1508 for 1.21.

Live tested on local/kvm.

(Review request: http://reviews.vapour.ws/r/832/)

281e85c


      Fixed lp:1416425 - backport for 1.21

9e2f308


      Merge pull request #1513 from dimitern/lp-1416425-bump-osext-rev-1.21

Fixed lp:1416425 - backport for 1.21

Backport of #1511 for 1.21.

(Review request: http://reviews.vapour.ws/r/835/)

a78642b


      Fix restore/azure

Azure was returning ErrNotBootstrapped from its StateServerInstances
method, which is not necessarily true. In the restore scenario run
by QA, Azure is bootstrapped and then the state server instance is
killed OOB. So, we change the StateServerInstances method to return
ErrNotBootstrapped instead. The "EnsureNotBootstrapped" code still
assumes that the environment is not bootstrapped if this error is
returned from StateServerInstances (i.e. the conversion moves up a
level).

The Azure provider will now also handle the scenario where there
are cloud services without deployments, returning ErrNoInstances
instead of ErrPartialInstances.

We also update the restore plugin to check for: ErrPartialInstances
(multiple state server instances created, some but not all alive),
and ErrNoInstances in the result of StateServerInstances.

Fixes https://bugs.launchpad.net/juju-core/+bug/1417178

70592db


      Address review comments

0d7d157


      Switch on errors.Cause(err)

33909a8


      Merge pull request #1529 from perrito666/1.21_fix_1417178

1.21 fix 1417178

Backport fix for 1417178

(Review request: http://reviews.vapour.ws/r/854/)

20a31b9


      Disabled networker worker and relevant tests - backport to 1.21

eaa92d0


      Merge pull request #1548 from dimitern/disable-networker-altogether-1.21

Disabled networker worker and relevant tests - backport to 1.21

Backport of #1493 for 1.21.

(Review request: http://reviews.vapour.ws/r/877/)

c2d7d40


      This includes updated branches for dependencies:

* Utils
* Syslog
* Testing

and fixes:

* 1416436
* 1416430
* 1416433

d3f2ab4


      Merge pull request #1553 from perrito666/fix_licencing

This includes updated branches for dependencies:

* Utils
* Syslog
* Testing

and fixes:

* 1416436
* 1416430
* 1416433

(Review request: http://reviews.vapour.ws/r/882/)

8966cad


      Charm revision updater dies on error

df1142c


      First error terminates as well

59469ab


      Test charmrevisionupdater dies on error

a0ef1b0

juju/juju

charmrevisionupdater retries on error #1564

voidspace commented Feb 9, 2015

gabriel-samfira and others added some commits Oct 29, 2014

jujubot and others added some commits Dec 4, 2014

voidspace closed this Feb 9, 2015

voidspace deleted the voidspace:1.21-charmrevup branch Oct 6, 2016

juju/juju

Join GitHub today

charmrevisionupdater retries on error #1564

Conversation

voidspace commented Feb 9, 2015

gabriel-samfira and others added some commits Oct 29, 2014

jujubot and others added some commits Dec 4, 2014

voidspace closed this Feb 9, 2015

voidspace deleted the voidspace:1.21-charmrevup branch Oct 6, 2016