Server plugin configuration settings always blank

Troubleshooting
1

Summary
I am currently writing a custom server plugin. I’ve updated the configuration struct to add a url string field, and I added a setting in plugin.json with a matching key.

Steps to reproduce
Clone GitHub - mattermost/mattermost-plugin-starter-template: Build scripts and templates for writing Mattermost plugins., then add a string field to the configuration as well as setting in plugin.json. Then make deploy to a Mattermost instance. Note that the value can be set in the plugin settings, but calls to plugin.getConfiguration().newValue are always empty.

This seems related to System Console Plugin Settings - #5 by agriesser

Observed Behavior
I noted that each time I attempted to set the plugin config value via the system console UI, I would get the following error:

2025-02-12 14:08:59 {"timestamp":"2025-02-12 20:08:59.577 Z","level":"warn","msg":"Unrecognized config permissions tag value.","caller":"api4/config.go:442","path":"/api/v4/config/patch","request_id":"rbwjquomk3n6ubpihejfmkookc","ip_addr":"172.17.0.1","user_id":"ke4yxq9f3in9f8opnw6b6ziyec","method":"PUT","tag_value":"sysconsole_write_*_read"}
2

Heya Jonathan! Welcome to the Mattermost forums. It looks like this might be due to a configuration or permissions issue, and I’m hoping someone here might have a workaround for ya soon. In the meantime, check out our Mattermost Plugin Configuration guide for more details.

1 Like
3

I’m using the Mattermost-Preview docker image, with the initial superuser uploading and configuring the plugin. Was the bug MVP reproduced?

4

Hi. After upgrading from 9.11 to 10.4.2, we also saw that config.json erases plugin information (keys, access points). Rolled back to 10.2.0 - the problem went away. But at the same time, we noticed that sending messages began to take longer. We tried to activate the enterpse version and build a cluster - but this did not bring success. Delays in sending continued to occur. After rolling back to 10.2.0 - the delays remained but became smaller.

1 Like
5

Just upgraded to 10.5.1 and this is still an issue.

6

So, I was able to reproduce the error in the test environment. What I did: activated the cluster mode. Cloned the virtual machine using vmware. Enabled the “notification” plugin /previously called “notification banner” on the production server. Started the cloned machine. The Mattermost service started on its own. After that, I added a node to the cluster. After restarting the service on the nodes, the config was reset. Some plugins were completely erased.
I’m trying to figure out what exactly caused this. Launching a cloned machine with identical machine-id or introducing nodes into the cluster with different config.json.
In this case, the cluster version does not seem to be fault-tolerant. If you have a cluster, the data will be erased on all nodes after restarting the application.