[BUG] Cannot List Systemwide Tasks

[jhunt]

Describe the bug
A clear and concise description of what the bug is.

To Reproduce
Steps to reproduce the behavior:

1. Run a SHIELD for a while
2. Run shield tasks as a system engineer (site-wide)
3. Pick a task UUID
4. Run shield task $uuid
5. Note failure.
6. Weep.

Expected behavior
I expected to see the log and metadata details of the task.

Screenshots

→  ./shield tasks -L | tail -n 5
9edb4cfa-a91f-4282-9a90-51cc441b8e49  test-store       done    system       2019-05-14 14:19:12-0400  2019-05-14 14:19:14-0400  2019-05-14 14:19:16-0400
e6dd83ae-880e-4840-b697-d37ffa4b3f39  backup           done    admin@local  2019-05-14 14:19:12-0400  2019-05-14 14:19:14-0400  2019-05-14 14:19:16-0400
48a69616-ed2a-4ff1-9609-bbfa469090cc  test-store       done    system       2019-05-14 14:19:11-0400  2019-05-14 14:19:14-0400  2019-05-14 14:19:16-0400
6208163e-28b6-4c13-83c6-e730df94f3d6  test-store       done    system       2019-05-14 14:19:11-0400  2019-05-14 14:19:14-0400  2019-05-14 14:19:16-0400
b4118328-5637-476a-9be7-20b6ceb147bd  agent-status     done    system       2019-05-14 14:19:03-0400  2019-05-14 14:19:04-0400  2019-05-14 14:19:04-0400

→  ./shield task b4118328-5637-476a-9be7-20b6ceb147bd
!!! No such task (no further diagnostic information available)

SHIELD versions (please complete the following information):

• SHIELD Core: 8.1+
• SHIELD CLI: same

Additional context
Found while trying to test short UUID support in new 8.2.x CLI.

[jhunt]

I think I found the root cause, since this unexpectedly worked this morning while I was tracking down some issues with a test-store task. Consider:

→  ./shield -c localdev tasks -s failed -L | grep test-store | tail -n 5
7414c483-f252-4f9c-8feb-5821059607bc  test-store  failed  system       2019-05-15 08:10:47-0400  2019-05-15 08:10:52-0400  2019-05-15 08:10:56-0400
98960a50-83bf-40ee-a13b-8675475c79d0  test-store  failed  system       2019-05-15 08:09:43-0400  2019-05-15 08:09:44-0400  2019-05-15 08:09:44-0400
07822665-a587-48ff-9e38-bbe504754c2b  test-store  failed  system       2019-05-15 08:09:03-0400  2019-05-15 08:09:03-0400  2019-05-15 08:09:07-0400
b3ba00d8-6054-42d6-b43f-fdbbde5c12b2  test-store  failed  system       2019-05-15 08:08:23-0400  2019-05-15 08:08:23-0400  2019-05-15 08:08:26-0400
0e87fc91-f142-46f6-9392-18e6b05f584e  test-store  failed  system       2019-05-15 08:07:43-0400  2019-05-15 08:07:48-0400  2019-05-15 08:07:53-0400

→  ./shield -c localdev task 7414c483-f252-4f9c-8feb-5821059607bc | head -n 5
UUID:       7414c483-f252-4f9c-8feb-5821059607bc
Owner:      system
Type:       test-store
Status:     failed

The difference between the failing case (as initially reported) and this working case is that the failing case was looking for an agent-status task, whereas the successful case is looking for a test-store task.

Need to determine, via -T if the CLI is always using /v2/tasks/:uuid for the non-tenant invocation (as all of these are), and if there are system-wide tasks that cannot be looked up that way.

Indeed, on the same system, with the same versions in play, a lookup of an agent-status task fails:

→  ./shield -c localdev tasks -L | grep agent-status | tail -n 5
bc60afc8-f73b-419f-88c0-e3a7ae7d15fd  agent-status     done    system       2019-05-15 08:09:43-0400  2019-05-15 08:09:43-0400  2019-05-15 08:09:44-0400
94ef18a5-01c0-46fb-8801-b31ff130c459  agent-status     done    system       2019-05-15 08:09:03-0400  2019-05-15 08:09:03-0400  2019-05-15 08:09:04-0400
c5ee0585-f3ec-4cfa-b970-1c54f40b85f2  agent-status     done    system       2019-05-15 08:08:23-0400  2019-05-15 08:08:23-0400  2019-05-15 08:08:26-0400
e7c39b77-a95f-47a7-96b2-68b9d77ee3de  agent-status     done    system       2019-05-15 08:07:43-0400  2019-05-15 08:07:48-0400  2019-05-15 08:07:51-0400
a65f09db-91bc-4cc6-8ea7-29028921ef45  agent-status     done    system       2019-05-15 08:07:08-0400  2019-05-15 08:07:08-0400  2019-05-15 08:07:08-0400

→  ./shield -c localdev task bc60afc8-f73b-419f-88c0-e3a7ae7d15fd | head -n 5
!!! No such task (no further diagnostic information available)

[jhunt]

Found this in the handler for /v2/tasks/:uuid:

        if task == nil || task.TenantUUID != db.GlobalTenantUUID {
            r.Fail(route.NotFound(err, "No such task"))
            return
        }

Sure enough, a quick SQL query for task types that have no tenant_uuid:

sqlite> SELECT op, COUNT(op) AS n FROM tasks WHERE tenant_uuid = '';
agent-status|34

I think that's the culprit. This will need a fix and a fixup.