do you have any best practices on upgrading the collector on Google Cloud? Basically I would try to do a rolling update of my compute engine and replace the startup script + application.conf. Since this is a critical part, do you have any advice on how to ensure proper rollout?
It’s best to do a rolling update, where you deprecate old nodes in favour of new nodes 1 by 1. If you notice any issues after each update, you roll back.
It’s also worth testing the configuration before deploying. You can use for example the stdout collector so you can send requests with cURL and verify that all looks good locally.
It might be necessary, eg if you plan to use custom paths or if you want to set a collector cookie on multiple domains from the same collector. (See the post linked below for specifics.)
For Safari ITP, what are the ideal settings?
It might be best to set SameSite=None. (See the post linked below for specifics.)
What would be the best way to validate a new config file?
The easiest way currently is to try and launch an stdout collector with the new config. If there are any issues, that will fail.
Is it sufficient to overwrite the config or do I have to restart the collector process?
You have to re-deploy the collector, using the new artefact. If you’re planning on using multiple domains, you’ll also have to configure the proper endpoints (which is not done via the hocon config and depends on which cloud you are).
We’ve tried to compile a list of known tricky issues when upgrading to Madara Rider. Please review and see if it answers your outstanding questions: