Arvados

I've also deployed a multi-node test cluster against latest development packages, and got the same error.
RailsAPI and Controller services are working, and manually running arv user current with the superuser token works OK from the shell node.

Manually running the same failing request succeeds:

admin@shell:~$ arv --format=uuid virtual_machine create --virtual-machine '{"hostname":"shell.xarv3.example.com" }'
xarv3-2x53u-b8qe0a7k7o7zug7

I think that I've found the behavior that causes this issue.

At cluster 'just-deployed' state¶

• The 'Not logged in' error happens when using the system root token
• Restarting controler and/or RailsAPI doesn't make the problem go away
• The users table has no records:
  

  xarv3_arvados=# select id,username,is_active from users;
   id | username | is_active 
  ----+----------+-----------
  (0 rows)
  

After logging in as an admin in workbench¶

• The problem no longer happens
• The users table contains both root and the admin user:
  

  xarv3_arvados=# select id,username,is_active from users;
   id | username | is_active 
  ----+----------+-----------
    1 |          | t
    2 | admin    | t
  (2 rows)
  

I could go ahead an manually do a binary search of which development package version started causing this, but first will confirm with the team if there's any suspicion on which branch could have introduced this new behavior.

Lucas Di Pentima wrote in #note-2:

Manually running the same failing request succeeds:

[...]

This succeeded because my manual test was done after logging in as the admin user in workbench for the first time.