all watcher api regression
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
juju-deployer |
Fix Released
|
Critical
|
Unassigned | ||
python-jujuclient |
Fix Released
|
Critical
|
Unassigned | ||
python-jujuclient (Ubuntu) |
Fix Released
|
High
|
Unassigned | ||
Trusty |
Fix Released
|
High
|
Unassigned |
Bug Description
It's back. It's also inconsistent.
Initially spotted by 'ev', https:/
ii juju-deployer 0.3.4-0ubuntu1 Deploy complex stacks of services using Juju
ii python-jujuclie 0.15-0ubuntu1 Python API client for juju
ii juju-core 1.17.3-0ubuntu1 Juju is devops distilled - client
daq-emitter is just an alias for the ganglia sub.
2014-02-25 15:26:17 Config specifies num units for subordinate: daq-emitter
Traceback (most recent call last):
File "/usr/bin/
load_
File "/usr/lib/
run()
File "/usr/lib/
importer.
File "/usr/lib/
self.
File "/usr/lib/
services=
File "/usr/lib/
services=
File "/usr/lib/
for change_set in self.watch:
File "/usr/lib/
return super(TimeoutWa
File "/usr/lib/
'Id': self.watcher_id})
File "/usr/lib/
raise EnvError(result)
jujuclient.
{ u'Error': u'state watcher was stopped', u'RequestId': 9, u'Response': { }}
>
Related branches
Changed in python-jujuclient (Ubuntu): | |
importance: | Undecided → High |
Changed in python-jujuclient: | |
importance: | Undecided → Critical |
status: | New → Fix Released |
Changed in python-jujuclient (Ubuntu Trusty): | |
status: | Fix Released → In Progress |
tags: | added: hs-arm64 |
Changed in juju-deployer: | |
status: | Confirmed → Fix Released |
So apparently mongo conn to state server goes awol sometimes, and everything has todo a forced reconnect. this has been happening frequently in juju 1.17.3 (at least the the third report i've had in the last week). There needs to be a deployer workaround at least for watches to detect this and auto-reconnect.