Does anyone know of a method to get rid of the bad node and the data that it is gathering?
I also do not know the correct way to achieve this.
But once I needed to rename one of the queues to follow the internal naming standard.
And I used the following queries on the provider node to prune _old_ queue:
SET search_path TO pgq_node;
DELETE FROM local_state WHERE queue_name='q-2rm';
DELETE FROM subscriber_info WHERE queue_name='q-2rm';
DELETE FROM node_info WHERE queue_name='q-2rm';
DELETE FROM node_location WHERE queue_name='q-2rm';
And restarted worker after that.
--
Victor Yegorov
Hi thanks for the answer. I've left this alone for a while and i've been looking into it in more detail today.
All the data is in the schema pgq. The other schema's are super small.
Also, the non-functional node is not in any of the tables in pgq_node.
But it is mentioned in pgq.consumer and linked with pgq.subscription.
When i remove the subscription records, nothing happens (worker and pgqd are running)
I was hoping hat some maintenance routine would remove the surplus data, but it doesn't.
I would delete the data myself but i can't figure out how to link that data to a consumer.
The event tables that contain all the data have no foreign keys. And i've tried to find mutual id's in those and other tables like pgq.tick but i haven't found anything of substance.
There are transaction id's in the event tables, i just hope that that is not the link because that would probably get seriously complicated to handle.
