[System Tests] Write tests for checking fuel upgrading feature

[Andrey Sledzinskiy]

Scenarious to automate will be:

Pre-condition for case 1:
System test successfully passed - environment was deployed and all checks passed

Case 1 - upgrade deployed environment (run if 'Upgrade_after_deployment' variable is set to true):
(These steps will be added to all tests after deploying environment and all checks)
1. Check there is no 'uuuuuupgrade_field' field by 'api/nodes'
2. Run upgrade (download tarball to master, extract files, run script with upgrade)
3. Check that cluster still exists and there is 'uuuuuupgrade_field' field by api/nodes, release version is changed
4. Run OSTF - all tests passed
5. Run 'docker images', 'docker ps' on master node - upgraded images and containers are shown, all containers are active
6. Check that - ls -l /etc/supervisord.d/ - current symlink is linked on /etc/supervisord.d/$upgraded_version
7. Check dockerctl shell works fine - it`s possible to “shell” into all of the containers
8. dockerctl destroy/stop/start works properly
9. check if there were no changes in permissions
10. Check that syslog gets HUP (analyze rsyslog)
11. For some tests we can add additional nodes, re-deploy cluster and run OSTF after that

Pre-condition for case 2:
Steps from Case 1 successfully passed

Case 2 - rollback upgraded deployed environment (run if 'Rollback' variable is set to true):
1. Execute next commands on master:
- rm /etc/supervisord.d/current
- ln -s /etc/supervisord.d/5.0/ /etc/supervisord.d/current
- /etc/init.d/supervisord stop
- docker stop $(docker ps -q)
- /etc/init.d/supervisord start
2. Check that cluster still exists, release version is changed to previous
3. Run OSTF - all tests passed
4. Run 'docker images', 'docker ps' on master node - images and containers with previous version are shown, all containers are active
5. Check that - ls -l /etc/supervisord.d/ - current symlink is linked on /etc/supervisord.d/$previous_version
6. Check dockerctl shell works fine - it`s possible to “shell” into all of the containers
7. dockerctl destroy/stop/start works properly
8. Check that syslog gets HUP (analyze rsyslog)
9. check if there were no changes in permissions

Pre-condition for case 3:
Pre-condition of system test is successfully passed, cluster is created and updated but not deployed

Case 3 - upgrade not deployed environment (run if 'Upgrade_before_deployment' variable is set to true):
1. Check there is no 'uuuuuupgrade_field' field by 'api/nodes'
2. Run upgrade (download tarball to master, extract files, run script with upgrade)
3. Check that cluster still exists and there is 'uuuuuupgrade_field' field by api/nodes, release version is changed
4. Run 'docker images', 'docker ps' on master node - upgraded images and containers are shown, all containers are active
5. Check that - ls -l /etc/supervisord.d/ - current symlink is linked on /etc/supervisord.d/$upgraded_version
6. Check dockerctl shell works fine - it`s possible to “shell” into all of the containers
7. dockerctl destroy/stop/start works properly
8. check if there were no changes in permissions
9. Check that syslog gets HUP (analyze rsyslog)
10. Add additional nodes for some tests

Pre-condition for case 4:
1,2 steps from Case 3 successfully passed

Case 4 - rollback upgraded not deployed environment (run if 'Rollback_before_deployment' variable is set to true):
1. Execute next commands on master:
- rm /etc/supervisord.d/current
- ln -s /etc/supervisord.d/5.0/ /etc/supervisord.d/current
- /etc/init.d/supervisord stop
- docker stop $(docker ps -q)
- /etc/init.d/supervisord start
2. Check that cluster still exists, release version is changed to previous
3. Run 'docker images', 'docker ps' on master node - images and containers with previous version are shown, all containers are active
4. Check that - ls -l /etc/supervisord.d/ - current symlink is linked on /etc/supervisord.d/$previous_version
5. Check dockerctl shell works fine - it`s possible to “shell” into all of the containers
6. dockerctl destroy/stop/start works properly
7. check if there were no changes in permissions
8. Check that syslog gets HUP (analyze rsyslog)

Pre-condition for case 5:
Case 3 successfully passed

Case 5 - rollback upgraded deployed environment (run if 'Rollback' variable is set to true):
1. Execute next commands on master:
- rm /etc/supervisord.d/current
- ln -s /etc/supervisord.d/5.0/ /etc/supervisord.d/current
- /etc/init.d/supervisord stop
- docker stop $(docker ps -q)
- /etc/init.d/supervisord start
2. Check that cluster still exists, release version is changed to previous
3. Run 'docker images', 'docker ps' on master node - images and containers with previous version are shown, all containers are active
4. Check that - ls -l /etc/supervisord.d/ - current symlink is linked on /etc/supervisord.d/$previous_version
5. Check dockerctl shell works fine - it`s possible to “shell” into all of the containers
6. dockerctl destroy/stop/start works properly
7. check if there were no changes in permissions
8. Check that syslog gets HUP (analyze rsyslog)

Case 6 - restart nodes after upgrade (probably should be added to test_failover.py):
1. Install master
2. Deploy simple cluster - simple, 1 controller, 1 compute, 1 cinder
3. Run upgrade (download tarball to master, extract files, run script with upgrade)
4. Restart all nodes
5. Check that nodes aren't bootstrapped

Case 7 - upgrade during deployment (probably should be added to test_failover.py):
1. Install master
2. Run deployment of ha cluster - 3 controller, 1 compute, 1 cinder
3. Run upgrade during deployment
4. Check that deployment stops
5. Check that cluster still exists and there is 'uuuuuupgrade_field' field by api/nodes, release version is changed
6. Run OSTF - all tests passed
7. Run 'docker images', 'docker ps' on master node - upgraded images and containers are shown, all containers are active
8. Check that - ls -l /etc/supervisord.d/ - current symlink is linked on /etc/supervisord.d/$upgraded_version
9. Check dockerctl shell works fine - it`s possible to “shell” into all of the containers
10. dockerctl destroy/stop/start works properly
11. check if there were no changes in permissions
12. Re-deploy cluster
13. Run OSTF

Case 8 - coexistence of environments with different fuel versions (probably should be added to test_failover.py):
1. Install master
2. Create simple cluster - simple, 1 controller, 1 compute, 1 cinder
3. Deploy cluster
4. Run OSTF
5. Upgrade fuel
6. Create new simple cluster with default values - simple, 1 controller, 1 compute, 1 cinder
7. Deploy new cluster
8. Run ostf for new cluster
9. Add 1 cinder node to old cluster and re-deploy
10. Run OSTF for old cluster

[Tatyanka (tatyana-leontovich)]

How do you plan to verify that upgrade is really happens ? you miss this point in all the scenarios

[Andrey Sledzinskiy (asledzinskiy)]

I'm going to check release version from api/version, version of docker containers. Isn't it enough?

[Dmitry Borodaenko (angdraug)]

Why is this a bug and not a blueprint?

[Andrey Sledzinskiy (asledzinskiy)]

Blueprint was created https://blueprints.launchpad.net/fuel/+spec/fuel-upgrade-system-tests
This bug will be to track progress of system tests writing

[Dmitry Borodaenko (angdraug)]

Progress of writing code for a blueprint should be tracked in the blueprint itself and linked gerrit reviews.

[Andrey Sledzinskiy (asledzinskiy)]

Ok, thanks, so I closed this bug as invalid