Guest User

Untitled

a guest
Apr 30th, 2015
263
0
Never
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
text 2.18 KB | None | 0 0
  1. Problems with full-stack testing:
  2.  
  3. 1) tools/configure_for_func_testing.sh fails to configure mysql for me. It says
  4. the password is wrong. This is a fresh Fedora 21 VM. I installed devstack before
  5. and set all passwords to '1'. Once I set the mysql root password to secretmysql
  6. the script could finish successfully. At this point I could run the functional DB
  7. tests (But the postgres tests were still skipped, I suspect this is wrong as well).
  8. I could also switch back the mysql root password back to the Devstack password
  9. and still have the DB functional tests finish successfully.
  10.  
  11. 2) Logging: It is assumed that /opt/stack/logs is the logs directory while in fact
  12. the logging directory is configurable in Devstack. That being said, /opt/stack is the
  13. default $DEST for devstack and the 'logs' subdirectory should exist after one's
  14. already used stack.sh. It was however kind of difficult to debug this: What happens
  15. if /opt/stack/logs doesn't exist is that the neutron-server process won't start
  16. and there's zero indication as to why. I think that the full-stack tests should make
  17. sure that the logs directory is there and writable by test user. If it's not there,
  18. create it. If it is but not writable, throw out an error and fail the test.
  19.  
  20. 3) Rabbit: Once I got the neutron-server process to start it failed to connect to Rabbit.
  21. I changed the password of the 'stackrabbit' user to '127.0.0.1'. This killed the devstack
  22. environment as it uses the password set in local.conf. Should we use a unique username / password?
  23.  
  24. 4) Concurrent testing: Unique vhost per test? This would require the startup of each test to
  25. add a vhost, and set_permissions on the test user to that vhost.
  26.  
  27. metadata_proxy is left after successful runs
  28.  
  29. I see different processes left after failed runs
  30.  
  31. Are br-int and br-ex being deleted after successful runs? Failed runs? (Spoiler: No)
  32.  
  33. After a while (Real helpful I know) rabbit completely jammed up. Every message
  34. send just froze, with no timeout. Queues were empty. Restarting the server or the VM
  35. did not help. unstack, clean, stack did not help. I did not try to stop_app, reset, start_app,
  36. add user, set permissions.
Advertisement
Add Comment
Please, Sign In to add comment