-
Notifications
You must be signed in to change notification settings - Fork 507
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Fix windows userdata #12859
Fix windows userdata #12859
Conversation
check-multi-juju, when running the unit tests on ubuntu, found: |
Since juju now only creates one service, the machine agent, we need to run that as a normal user.
7eef4ac
to
abee5b2
Compare
!!build!! |
@gabriel-samfira please add QA steps to the PR description. |
During QA with 2.9, the windows machine gets stuck in pending. I'm using OpenStack and the Windows 2012 R2 eval KVM image you pointed me at. The machine is active, but the juju agent never connects to the controller. I'm I building the windows agent binaries with: I can successfully add a Windows 2012 R2 eval machine to a 2.8.10 controller on the same OpenStack. |
Try the following: make release-install
GOOS=windows make release-install Generate simplestreams for agents: mkdir -p $HOME/simplestreams/tools/proposed
cp $GOPATH/bin/jujud $HOME/simplestreams/tools/proposed
cp $GOPATH/bin/windows_amd64/jujud.exe $HOME/simplestreams/tools/proposed
cd $HOME/simplestreams/tools/proposed
tar czf juju-2.9.1-ubuntu-amd64.tgz jujud
tar czf juju-2.9.1-windows-amd64.tgz jujud.exe
rm jujud jujud.exe
cd ~
juju-metadata generate-agents -d $HOME/simplestreams --stream proposed Copy the tools folder to a web server that is accessible to your openstack environment, then bootstrap using: juju bootstrap --debug --verbose --show-log \
--config image-metadata-url="http://YOUR_SERVER_IP/images/" \
--config agent-metadata-url="http://YOUR_SERVER_IP/tools/" \
--config agent-stream=proposed oss oss-ctrl Once bootstrapped, make sure the checksum of the If it still fails, please post the log from: C:\Program Files\Cloudbase Solutions\Cloudbase-Init\log\cloudbase-init.log Also, on the windows instance, you can check the existance of the windows service representing the agent: Get-Service juju* To get info about the service: sc.exe qc jujud-machine-0 |
QA worked today. Maybe it took the OpenStack I was using to be upgraded? |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
QA good. Thank you for the contribution.
|
#13050 Merge 2.9 #12859 Fix windows userdata #12986 Prevent contention error when switching charms that drop peer relations #13001 Charmhub: Improve download logging on failure #13003 Only run upgrade-charm hook once when k8s charm upgraded #13004 Update Pebble: enable "pebble run --verbose" logging and starting services as different user/group #13007 Lp1920820 dashboard proxy support #13002 CLI: Fix panic in upgrade-series with older controllers #13036 Use mongo storage for external identity macaroon keys #13040 Integration tests: Extend timeout #13033 Remove stand-alone unit agent concerns #13037 Handle SIGTERM and stop hook on sidecar deploys #13038 Use the address CIDRs when populating machine addresses in the instance-poller #13039 Fixes error with ensure ClusterRole #13041 Update apt mirrors on all nodes when model value changes #13043 Integration tests: Manual test charm selection #13045 Fix race in caasunitterminationworker tests #13044 Do not attempt downgrades in the upgrade-steps worker #13046 Fix display of scale and workload version in status #13048 Data race state/logdb tests #13049 Report service address for sidecar k8s charms #13031 vsphere values for show-cloud --include-config ``` # Conflicts: # apiserver/facades/client/application/update_series_mocks_test.go # apiserver/facades/client/bundle/bundle.go # apiserver/facades/client/machinemanager/types_mock_test.go # cmd/juju/application/bundle/bundle.go # cmd/juju/application/bundle/bundle_test.go # cmd/juju/application/deployer/bundlehandler.go # cmd/juju/application/diffbundle.go # cmd/juju/dashboard/dashboard.go # featuretests/upgrade_test.go # go.mod # go.sum # scripts/win-installer/setup.iss # snap/snapcraft.yaml # state/action_test.go # version/version.go # worker/uniter/resolver/loop.go ``` ## QA steps See PRs
This fixes an embarrassing bug of my own making, introduced in commit: 942456e
where I omitted to remove the corresponding parameters to
fmt.Sprintf
after modifying the string it was formatting.I added a test for
InstallCommands()
as well as fixing some tests for password change that seem to have not been caught by the CI.Update:
Given that unit agents seem to be gone now, and there is only one service being created on a machine, the machine agent is now run under the local jujud user on windows again.
A followup PR will be created at a later date that will remove code that was used to maintain unit agents on Windows.
QA steps
Wait for machine to boot.
juju run --machine 0 "sc.exe qc jujud-machine-0"
Check that
SERVICE_START_NAME
is.\jujud
.