configured inputs do not persist about graylog2-images

[mariussturm]

Could you please pull the latest image and provide this parameter when starting the container: -e GRAYLOG_NODE_ID=30833ff4-f14c-4641-8001-0866ac439552 (randomize the ID string!) This should now set the node ID to defined value. When the container is recreated on top of old data you can re-use the same ID. This should persist even inputs.

[mariussturm]

Yes the inputs are stored in MongoDB. The problem is that they are associated through the servers node-id with a defined server instance. So every time you start a complete new container based on an old database, you get a new server node-id and the old inputs don't belong to this server instance anymore. The idea was to just stop and start containers. In this way the node-id should be the same.

[MrMarvin]

Ah, I see now.
So when using global inputs instead, they persist for every container instance I assume?!
Although it's kind of is obvious, I feel like this should be documented in the docker/README.md.

[mariussturm]

Global means that a cluster of Graylog servers will all start the same input. But they are still bound to a node-id. Yep, will drop a line in the docs.

[mariussturm]

Closing now because no real issue. Documentation is added.

[tzz]

@mariussturm does specifying a container name with --name mycontainer set the node ID persistently? Or is it tied to something else?

[mariussturm]

do you mean docker run --name ...? That is not used inside of the container.

[tzz]

I actually meant docker run --hostname=myname, sorry. That sets the hostname as far as all the apps inside the container are concerned so maybe it can be used to persist the settings.

[mariussturm]

I think the cleanest solution would be to introduce a new container option for it. So people could provide the node-id through another environment variable.

[tzz]

Works for me :) I just thought setting the hostname (which typically is a random string if you don't provide it explicitly) might be a nice way to provide the node ID. It would be nice to use the same node ID for ES as well, actually, so you can identify it easily in the ES logs.

[sheavner]

I grabbed the nodeid from the currenlty running node and passed that in after a container destroy and restart. There is now a node running with that old id.

I can't verify that the inputs are working, because I've setup mine to be global (started on all nodes) so they always come back even without the persisted node id.

I have a different issue. I have one custom user. That user exists, but I can't login and get invalid credentials. After each restart I have to go to System > Users and update the password for this user. And then things are fine. The persistent GRAYLOG_NODE_ID didn't fix this.

[tzz]

@sheavner how do you grab the node ID of a running node? I didn't find that in the docs and it's not obvious from the Graylog UI. Thanks!

The invalid user password is definitely inconvenient but not as bad as losing data. I hope it can be resolved.

[mariussturm]

Ok, thanks for the feedback! I think we have to expose the password_secret as well. Will take a look soon.

[sheavner]

@tzz If there's only one node, it's obvious from the system messages on the System Overview page at /system?page=0. I think I found mine on /system/nodes which shows the short hash + nodename. If you click on the node name, you can get the full nodeid from the url as you're viewing the node details.

[tzz]

Thanks @mariussturm I found it now. I knew the short version (which I saw elsewhere) was not enough but didn't notice the node detail page. I'll wait for the password secret but otherwise this persistence is a great feature.

[mariussturm]

Could you please pull the latest image and add this parameter to set the server password salt: -e GRAYLOG_SERVER_SECRET=somerandomstring
Next time you spin up the container with the same secret the user password should work.

[sheavner]

@mariussturm My non-admin user password did persist across a test restart of the server after the pull. Thanks for fixing this.

[mariussturm]

Thanks for testing!