Windows 10 Client Distributions dashboard hasn't updated in weeks

RESOLVED FIXED

Status

Data Platform and Tools
Redash (STMO)
P1
normal
RESOLVED FIXED
3 months ago
2 months ago

People

(Reporter: RyanVM, Assigned: harter)

Tracking

Details

(URL)

(Reporter)

Description

3 months ago
Looking at the dashboard, the charts at the top say they haven't been updated in 17 days and ones below for over a month. Is something stuck? When I try to manually refresh, I get an "Error running query: line 3:6: Failed analyzing stored view  'hive.default.build_distribution': Access Denied: Cannot create view default.windows_10_aggregate" error.

Also, the column ordering on the "Release Channel Distribution" chart appears to have gotten mixed up. Would be great if we could fix that too.
(Reporter)

Comment 1

3 months ago
Triage note - this is one I do look at regularly, so I would appreciate it being unbroken in a relatively short timeframe :)
(Reporter)

Comment 2

3 months ago
Ryan, do you have cycles to look into this or can you redirect to someone who does?
Flags: needinfo?(rharter)
(Assignee)

Updated

3 months ago
Assignee: nobody → rharter
(Assignee)

Comment 3

3 months ago
Looks like virtual views may no longer be supported in STMO? Blake, do you know of any recent changes that may have caused this permissions issue?

Looks like the windows_10_aggregates view [0] runs fine but downstream queries using the view fail [1].

[0] https://sql.telemetry.mozilla.org/queries/776/source
[1] https://sql.telemetry.mozilla.org/queries/817/source
Flags: needinfo?(rharter) → needinfo?(bimsland)
This appears to be due to a permissions change with how we authorize access to particular hive tables, particularly that creating a view that references another view requires the GRANT_SELECT permission [1] which was not set. This has been added to the config and I was able to successfully run the query that refreshed the build_distribution view, currently running the other downstream queries to verify it works all the way through.

[1] https://prestodb.io/docs/0.180/connector/hive-security.html#table-rules
Flags: needinfo?(bimsland)
(Reporter)

Comment 5

2 months ago
Looks like things are mostly working now, except the patch adoption charts at the bottom are still broken with "Error running query: type of symbol 'expr_10471' is expected to be bigint, but the actual type is varchar" messages.
Flags: needinfo?(rharter)
(Assignee)

Updated

2 months ago
Flags: needinfo?(rharter)
Priority: -- → P1
(Assignee)

Updated

2 months ago
See Also: → bug 1399532
(Assignee)

Comment 6

2 months ago
This should now be fixed. There was a broken CAST in an upstream query due to a new type of errant data. Additionally, Presto now appears to have issues with a (now fixed) naming collision that was affecting line 3 of this query [1].

I filed Bug 1399532 to make this more maintainable.

[1] https://sql.telemetry.mozilla.org/queries/779/source#table
Status: NEW → RESOLVED
Last Resolved: 2 months ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.