Make datasource UID editable in GUI

[XANi]

Previously, as long as 2 grafana instances (say dev and prod, or dc1 and dc2) had identically named data sources it was relatively straightforward to either export/import dashboards between them manually or automate that.

That was hacky (it relied on 2 instances having same name for data source), and caused problems with renaming ( #1664 #33817 ) but generally worked.

But in recent version exported dashboard JSON now contains UID.

Now name of the data source is user-controllable but UID is not, which means only way for new instance to accept dashboards exported from old instance is either:

• set up provisioning with same UIDs (which is fine solution if you're already automated your deployments but otherwise "too late")
• editing database to match that UID (and as I understand break any existing dashboards in the database)
• editing dashboard JSONs to use legacy by-name method

It would be useful if there was an option to change the datasouce UID, either in web config (altho that would possibly require migrating existing ones), or to have it as option during export.

[sojuz151]

What about making the data source into a variable with the instance name filter machtich only one name?

[XANi]

Dunno. "Export for sharing externally" already kinda does 90% of the work here as it replaces UID with "${DS_INFLUXDB}" but that doesn't help if your instance has more than one influxdb database used.

[xkilian]

This is also a huge issue for us, as we have air-gapped networks that have similar dashboards, and we can no longer import/export dashboards between the environments. We need a solution.

[XANi]

@xkilian the "hack" is to just make sure all of them have same UID (either changing one in DB or deploying one via provisioning) but it's definitely less trivial solution

[rafaelpirolla]

PLEASE TELL ME IF THIS WILL BREAK ANY OTHER PART OF CONFIG

Example:
get the uid by checking the datasource in a json of a current dashboard

systemcl stop grafana-server
sqlite3 /var/lib/grafana/grafana.db
>select * from data_source;
1|1|1|prometheus|Prometheus|...|...|...|...|...|...|...|...|...|...|...|...|...|...|...|000000001
...
>update data_source set uid="<uid here>" where name="Prometheus";

schema:

>.schema
(...)
CREATE TABLE `data_source` (
`id` INTEGER PRIMARY KEY AUTOINCREMENT NOT NULL
, `org_id` INTEGER NOT NULL
, `version` INTEGER NOT NULL
, `type` TEXT NOT NULL
, `name` TEXT NOT NULL
, `access` TEXT NOT NULL
, `url` TEXT NOT NULL
, `password` TEXT NULL
, `user` TEXT NULL
, `database` TEXT NULL
, `basic_auth` INTEGER NOT NULL
, `basic_auth_user` TEXT NULL
, `basic_auth_password` TEXT NULL
, `is_default` INTEGER NOT NULL
, `json_data` TEXT NULL
, `created` DATETIME NOT NULL
, `updated` DATETIME NOT NULL
, `with_credentials` INTEGER NOT NULL DEFAULT 0, `secure_json_data` TEXT NULL, `read_only` INTEGER NULL, `uid` TEXT NOT NULL DEFAULT 0);
(...)

[shaunmansell]

Anyone know exactly what Grafana version brought in the change to the JSON spec to now include datasource UIDs?

I'm looking for the highest version I could roll back to where I won't hit issues in our cross environment dashboard provisioning that has been impacted by UIDs in the dashboard JSON

Edit: Answering my own question. It came out in 8.3.0 via #33817

[laumarf3]

It would be great to have the possibility of not exporting the UID, but also the database name!

[youssefbenfarhat]

Hello, I'm facing the same error.

Did you find a solution for this ?

[XANi]

@youssefbenfarhat we just changed our datasources to have same UID between sites (via provisioning) but that's a half measure

[youssefbenfarhat]

What do you mean by a half measure ?

[XANi]

Well, we have a bunch of same instance using mostly same graphs but serving different sites. Wouldn't work too well if environment wasn't so homogenous

[gplechuck]

Hey @youssefbenfarhat , @XANi

Stumbled across this post this morning after encountering a similar issue provisioning a dashboard across sites (with different datasource UIDs) .

An alternative workaround -

• export the dashboard JSON, click 'View JSON' and copy that
• change any occurrences of the datasource reference to the older style format , eg

"datasource": {
  "type": "prometheus",
  "uid": "WAwoBA8Gk"
},

to this -
"datasource": "prometheus",

This allowed us to share dashboards between sites where the datasources were similarly named (eg prometheus), but had different UIDs.

Hope that helps .

[dkrut]

Thanks! It helps for me. I'm using influxdb

[mozai]

Ran into this today. Made a new panel in the test env, exported the dashboard JSON to our helmchart to deploy across the enterprise. Next morning, emails from confused users because of "datasource not found" because even though the datasource definition is identical (same helm chart) across all clusters, the generated UID is different between test and prod. If I could enforce a uid string in a datasource definition, like we can with dashboard definitions, that might help.

[XANi]

The "fix" for it currently is provisioning datasources via JSON files rather than via UI and keeping consistent UID

[gelicia]

Hello, as you may have heard, we are transitioning away from using discussions to discuss feature requests. We are migrating this discussion to an issue and closing the discussion. The issue is #82962. Feel free to continue the discussion around this there. Thank you!