grafana-operator: [Bug] Installing plugins fails
Describe the bug
Installing a plugin via the admin console fails. There is no specific error message in the UI, but the logfile says Failed to install plugin" error="failed to create temporary file: open /tmp/1141914888.zip: read-only file system
. I suppose every K8s has a read-only filesystem. I haven’t found any other way to install plugins, e. g. setting GF_INSTALL_PLUGINS
fails for the same reason.
Version 5.0.0-rc0
To Reproduce Steps to reproduce the behavior:
see above
Expected behavior I would expect it to be possible to install a plugin with the Operator. The docs do not provide any insight on this, except for a situation where a dashboard-plugin can be installed with a dashboard. But in my case it is about a datasource plugin.
Runtime (please complete the following information):
- OS: Linux
- Environment: Kubernetes (k3s)
About this issue
- Original URL
- State: closed
- Created 10 months ago
- Comments: 22
The operator does not install any plugins. Its only job in this context is to properly generate
GF_INSTALL_PLUGINS
env based onGrafanaDashboard
/GrafanaDatasource
CRs. Then, each Grafana instance installs the plugins during its boot up process.When you click on something in Grafana UI, the operator is not aware of that, because you’re sending API requests directly to a Grafana instance, not to the operator. So, it’s not the operator’s fault that the result you expect doesn’t match the actual state. IaC principles are around declaring a desired state (in CRs) and not changing anything manually. I believe it’s not something worth documenting, because the whole Kubernetes eco-system is built like that.
If you want to somehow limit plugin installation after Grafana boots up, you need to raise a feature request / PR in Grafana repository. It’s outside of our controls.