19.09.1 does not load custom module
kappac opened this issue · comments
I'm trying to update my setup from 19.05 to 19.09.1, but get an error:
[critical] <0.110.0>@ejabberd_app:start:71 Failed to start ejabberd application: Invalid value of option auth_method->1: unsupported database: something.
Everything works as expected with 19.05.
Should be auth modules deployed in a different way or is it a bug?
Can you please start ejabberd with some built-in authentication, let's say mnesia, and then change the config to use your auth method and reload the config. Do you see the same error?
We use docker setup. I'm not sure reloading will change something.
Is there any way to log modules' loading?
I'm not sure reloading will change something.
Okay, keep being unsure.
When you're ready to do what I ask to do to find the problem let me know.
There is no errors if mnesia in auth_method is used.
...
auth_method:
- mnesia
...
Docker image is configured with additional scripts to start ejabberd, so it's impossible to update config inside the container and restart ejabberd.
With custom auth method the error is: [critical] <0.110.0>@ejabberd_app:start:71 Failed to start ejabberd application: Invalid value of option auth_method->1: unsupported database: something
Another one error with mnesia in auth_method:
[critical] <0.110.0>@ejabberd_app:start:71 Failed to start ejabberd application: Invalid value of option modules: unknown ejabberd module: ejabberd_auth_nextiva. Did you mean mod_push_mnesia?
ejabberd_auth_nextiva is configured in modules section.
I'm not talking about start, I'm talking about configuration reload, e.g. via ejabberdctl reload-config.
It's impossible in current setup, I'm trying to do another setup to do what you've requested.
Okay, let's try another way: start ejabberd with mnesia auth method, then attach to running ejabberd via ejabberdctl debug and type in the console:
code:ensure_loaded(ejabberd_auth_nextiva).
(don't forget the trailing dot).
What is the output?
It's impossible in current setup
Well, that's a bad setup then.
I honestly think all the problems come from the docker image itself.
It's not bad, it's implemented according to business requirements. Anyway, I've a way to reload ejabberd.
Will update thread in several minutes.
With 19.05 there is no any problems.
It's not bad, it's implemented according to business requirements.
So, service interruption just for the configuration change is a part of business requirement?
Service does not interrupt if ejabberd is 19.05, that's why I've opened an issue.
I've got this error after ejabberdctl reload_config: [error] <0.670.0>@ejabberd_config:reload:114 Configuration reload aborted: Configuration error: missing required option: host.
In the Configuring ejabberd I'm not able to find description of this option, except modules description.
I told you to try code:ensure_loaded(), see above.
Whatever, just put your ejabberd_auth_nextiva.beam to other ejabberd beams. This will work for sure.
ejabberdctl debug hangs up, while other ejabberdctl commands work fine.
Thank you!