better communication of database request "failures"
Bug #1797346 reported by
Junien F
This bug affects 1 person
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| pgsql Interface for charms.reactive |
Triaged
|
Medium
|
Unassigned | ||
Bug Description
Hi,
I just had a weird bug where I deployed a unit ("sentry" charm) that requested a database without extension, then the charm got upgraded and other units added, so these other units also requested a database but this time with the "citext" extension. And they never got a database back (ie the "postgresql.
Could we maybe improve the experience here ? Not sure how to be quite honest...
Thanks
Revision history for this message
| Stuart Bishop (stub) wrote | #1 |
| Changed in interface-pgsql: | |
| status: | New → Triaged |
| importance: | Undecided → Medium |
To post a comment you must log in.
The clients were inconsistent with their database requests, with only one unit requesting the 'citext' extension. This unit never considered the database available, because the server had not installed the requested extension yet, because the server only looks at an arbitrary unit and trusts that they will converge (a common assumption).
The interface has access to layer:status now, which we can use to populate the status with informative messages during relation handshaking. We can also log messages (master is not yet available because [...])