I don't see any successes for multi-series-charm so I'm not sure that there
is actually a cause from this patch.
The failure around 'agent-lost' or
'agent-is-not-communicating-with-the-server' could, indeed be caused by the
change from this patch, as it is specifically dealing with changes around
detecting whether agents are alive and communicating or not.
I haven't managed to actually run the test yet, but just doing "bootstrap
&& deploy" I don't see the result. I've been trying with Xenial and LXD, I
may try with Trusty to see if it makes a difference. (Its possible that
Mongo 2.4 on Trusty doesn't support the $bit operation as we think it does,
though the test suite was passing.)
On Mon, Oct 2, 2017 at 12:07 PM, John Meinel <email address hidden> wrote:
> Given the need to run against $ENV, I'm guessing this expects to have a
> controller already bootstrapped? There shouldn't be anything in the patch
> that changes agents (vs possibly changing how controllers operate).
>
> I'm missing something in how to run the tests manually. If I try:
>
> ./assess_storage lxd ~/dev/go/bin/juju ~/dev/tmp/juju_test
>
> Then it warns me that:
> /home/jameinel/dev/go/src/github.com/juju/juju/
> acceptancetests/utility.py:88: UserWarning: Not a directory
> /home/jameinel/dev/tmp/jujutest
> warnings.warn('Not a directory {}'.format(maybe_dir))
> and fails with:
> File "/home/jameinel/dev/go/src/github.com/juju/juju/
> acceptancetests/jujupy/configuration.py", line 37, in
> get_selected_environment
> 'Environment "{}" does not exist.'.format(selected))
> jujupy.configuration.NoSuchEnvironment: Environment "lxd" does not exist.
>
> However, if I first do "juju bootstrap lxd" I then get:
> /home/jameinel/dev/go/src/github.com/juju/juju/
> acceptancetests/utility.py:94: UserWarning: Directory
> '/home/jameinel/dev/tmp/jujutest' has existing contents.
> 'Directory {!r} has existing contents.'.format(maybe_dir))
>
> and a final failure of:
> jujupy.configuration.NoSuchEnvironment: Environment "lxd" does not exist.
>
> Now, trying to look at the code, I see it trying to do:
>
> def get_juju_home():
> home = os.environ.get('JUJU_HOME')
> if home is None:
> home = os.path.join(os.environ.get('HOME'), '.juju')
> return home
> ...
> def get_environments_path(juju_home):
> return os.path.join(juju_home, 'environments.yaml')
> ...
> def get_environments():
> """Return the environments for juju."""
> home = get_juju_home()
> with open(get_environments_path(home)) as env:
> return yaml.safe_load(env)['environments']
>
> So it seems that this code assumes you have a leftover Juju-1.x setup
> laying around that has a matching environments.yaml file?
>
>
> On Mon, Oct 2, 2017 at 11:37 AM, John Meinel <email address hidden>
> wrote:
>
>> Note that you have to be logged in to see the 'charm-storage' job.
>> Otherwise those links come back as 404.
>>
>> I'm investigating this now.
>>
>> John
>>
>> On Thu, Sep 28, 2017 at 7:50 PM, Burton Swan <email address hidden>
>> wrote:
>>
>>> ** Description changed:
>>>
>>> - Juju version: 2.3-alpha1
>>> + Juju version: 2.3-alpha1 and 2.2 series.
>>>
>>> PR7868 is highly likely the cause of this issue:
>>> https://github.com/juju/juju/pull/7868
>>>
>>> Using charm-storage test as an example:
>>> http://ci.jujucharms.com/job/charm-storage/
>>>
>>> Re-run the last successful job, revision build 5743, commit a906cf9,
>>> works well:
>>> http://ci.jujucharms.com/job/charm-storage/2027/
>>>
>>> Then re-run the first failed job, revision build 5745, commit 774f2f0,
>>> failed again:
>>> http://ci.jujucharms.com/job/charm-storage/2028/
>>> 2017-09-28 14:11:49 ERROR ('0', 'agent is not communicating with the
>>> server', 1149.777357)
>>>
>>> commit 774f2f0 also triggered other failures, e.g.:
>>> http://ci.jujucharms.com/job/multi-series-charm/
>>>
>>> This issue can also be manually reproduced on the latest Juju
>>> 2.3-alpha1 by run:
>>> $ ./assess_storage.py $ENV $JUJU_BIN $JUJU_DATA
>>>
>>> ** Description changed:
>>>
>>> Juju version: 2.3-alpha1 and 2.2 series.
>>>
>>> PR7868 is highly likely the cause of this issue:
>>> https://github.com/juju/juju/pull/7868
>>> +
>>> + On Juju 2.2 series it is PR7863:
>>> + https://github.com/juju/juju/pull/7863
>>>
>>> Using charm-storage test as an example:
>>> http://ci.jujucharms.com/job/charm-storage/
>>>
>>> Re-run the last successful job, revision build 5743, commit a906cf9,
>>> works well:
>>> http://ci.jujucharms.com/job/charm-storage/2027/
>>>
>>> Then re-run the first failed job, revision build 5745, commit 774f2f0,
>>> failed again:
>>> http://ci.jujucharms.com/job/charm-storage/2028/
>>> 2017-09-28 14:11:49 ERROR ('0', 'agent is not communicating with the
>>> server', 1149.777357)
>>>
>>> commit 774f2f0 also triggered other failures, e.g.:
>>> http://ci.jujucharms.com/job/multi-series-charm/
>>>
>>> This issue can also be manually reproduced on the latest Juju
>>> 2.3-alpha1 by run:
>>> $ ./assess_storage.py $ENV $JUJU_BIN $JUJU_DATA
>>>
>>> --
>>> You received this bug notification because you are subscribed to the bug
>>> report.
>>> https://bugs.launchpad.net/bugs/1720155
>>>
>>> Title:
>>> Juju agent is not communicating with the server.
>>>
>>> To manage notifications about this bug go to:
>>> https://bugs.launchpad.net/juju/+bug/1720155/+subscriptions
>>>
>>
>>
>
I don't see any successes for multi-series-charm so I'm not sure that there
is actually a cause from this patch.
The failure around 'agent-lost' or is-not- communicating- with-the- server' could, indeed be caused by the
'agent-
change from this patch, as it is specifically dealing with changes around
detecting whether agents are alive and communicating or not.
I haven't managed to actually run the test yet, but just doing "bootstrap
&& deploy" I don't see the result. I've been trying with Xenial and LXD, I
may try with Trusty to see if it makes a difference. (Its possible that
Mongo 2.4 on Trusty doesn't support the $bit operation as we think it does,
though the test suite was passing.)
On Mon, Oct 2, 2017 at 12:07 PM, John Meinel <email address hidden> wrote:
> Given the need to run against $ENV, I'm guessing this expects to have a dev/go/ src/github. com/juju/ juju/ /utility. py:88: UserWarning: Not a directory dev/tmp/ jujutest maybe_dir) ) jameinel/ dev/go/ src/github. com/juju/ juju/ /jujupy/ configuration. py", line 37, in environment .format( selected) ) configuration. NoSuchEnvironme nt: Environment "lxd" does not exist. dev/go/ src/github. com/juju/ juju/ /utility. py:94: UserWarning: Directory jameinel/ dev/tmp/ jujutest' has existing contents. '.format( maybe_dir) ) configuration. NoSuchEnvironme nt: Environment "lxd" does not exist. get('JUJU_ HOME') join(os. environ. get('HOME' ), '.juju') s_path( juju_home) : join(juju_ home, 'environments. yaml') environments_ path(home) ) as env: load(env) ['environments' ] /github. com/juju/ juju/pull/ 7868 ci.jujucharms. com/job/ charm-storage/ ci.jujucharms. com/job/ charm-storage/ 2027/ ci.jujucharms. com/job/ charm-storage/ 2028/ ci.jujucharms. com/job/ multi-series- charm/ /github. com/juju/ juju/pull/ 7868 /github. com/juju/ juju/pull/ 7863 ci.jujucharms. com/job/ charm-storage/ ci.jujucharms. com/job/ charm-storage/ 2027/ ci.jujucharms. com/job/ charm-storage/ 2028/ ci.jujucharms. com/job/ multi-series- charm/ /bugs.launchpad .net/bugs/ 1720155 /bugs.launchpad .net/juju/ +bug/1720155/ +subscriptions
> controller already bootstrapped? There shouldn't be anything in the patch
> that changes agents (vs possibly changing how controllers operate).
>
> I'm missing something in how to run the tests manually. If I try:
>
> ./assess_storage lxd ~/dev/go/bin/juju ~/dev/tmp/juju_test
>
> Then it warns me that:
> /home/jameinel/
> acceptancetests
> /home/jameinel/
> warnings.warn('Not a directory {}'.format(
> and fails with:
> File "/home/
> acceptancetests
> get_selected_
> 'Environment "{}" does not exist.'
> jujupy.
>
> However, if I first do "juju bootstrap lxd" I then get:
> /home/jameinel/
> acceptancetests
> '/home/
> 'Directory {!r} has existing contents.
>
> and a final failure of:
> jujupy.
>
> Now, trying to look at the code, I see it trying to do:
>
> def get_juju_home():
> home = os.environ.
> if home is None:
> home = os.path.
> return home
> ...
> def get_environment
> return os.path.
> ...
> def get_environments():
> """Return the environments for juju."""
> home = get_juju_home()
> with open(get_
> return yaml.safe_
>
> So it seems that this code assumes you have a leftover Juju-1.x setup
> laying around that has a matching environments.yaml file?
>
>
> On Mon, Oct 2, 2017 at 11:37 AM, John Meinel <email address hidden>
> wrote:
>
>> Note that you have to be logged in to see the 'charm-storage' job.
>> Otherwise those links come back as 404.
>>
>> I'm investigating this now.
>>
>> John
>>
>> On Thu, Sep 28, 2017 at 7:50 PM, Burton Swan <email address hidden>
>> wrote:
>>
>>> ** Description changed:
>>>
>>> - Juju version: 2.3-alpha1
>>> + Juju version: 2.3-alpha1 and 2.2 series.
>>>
>>> PR7868 is highly likely the cause of this issue:
>>> https:/
>>>
>>> Using charm-storage test as an example:
>>> http://
>>>
>>> Re-run the last successful job, revision build 5743, commit a906cf9,
>>> works well:
>>> http://
>>>
>>> Then re-run the first failed job, revision build 5745, commit 774f2f0,
>>> failed again:
>>> http://
>>> 2017-09-28 14:11:49 ERROR ('0', 'agent is not communicating with the
>>> server', 1149.777357)
>>>
>>> commit 774f2f0 also triggered other failures, e.g.:
>>> http://
>>>
>>> This issue can also be manually reproduced on the latest Juju
>>> 2.3-alpha1 by run:
>>> $ ./assess_storage.py $ENV $JUJU_BIN $JUJU_DATA
>>>
>>> ** Description changed:
>>>
>>> Juju version: 2.3-alpha1 and 2.2 series.
>>>
>>> PR7868 is highly likely the cause of this issue:
>>> https:/
>>> +
>>> + On Juju 2.2 series it is PR7863:
>>> + https:/
>>>
>>> Using charm-storage test as an example:
>>> http://
>>>
>>> Re-run the last successful job, revision build 5743, commit a906cf9,
>>> works well:
>>> http://
>>>
>>> Then re-run the first failed job, revision build 5745, commit 774f2f0,
>>> failed again:
>>> http://
>>> 2017-09-28 14:11:49 ERROR ('0', 'agent is not communicating with the
>>> server', 1149.777357)
>>>
>>> commit 774f2f0 also triggered other failures, e.g.:
>>> http://
>>>
>>> This issue can also be manually reproduced on the latest Juju
>>> 2.3-alpha1 by run:
>>> $ ./assess_storage.py $ENV $JUJU_BIN $JUJU_DATA
>>>
>>> --
>>> You received this bug notification because you are subscribed to the bug
>>> report.
>>> https:/
>>>
>>> Title:
>>> Juju agent is not communicating with the server.
>>>
>>> To manage notifications about this bug go to:
>>> https:/
>>>
>>
>>
>