Alexander Brand

I haven't thought through the implications, but I think we should be able to swallow the "does not exist" error when we are trying to delete, instead of raising it...

These charts are interesting. We can see when the namespace is deleted and all services are gone, but the discoverer is still trying to delete them, and thus accumulating errors....

Memory usage of discoverer: 

Yeah, I actually think we need both... It doesn't look like we have a max retry on the queue item. I'll pick this up and look into it more.

Now that we know we were not actually retrying before #141, the issue title was somewhat misleading. What was actually going on is that the namespace is deleted from under...

The more I think about this, the more I am second guessing whether we should do it. It seems like the deletion of a namespace with thousands of services is...

Oh! The refresh not only updates the resources in the cache, but also adds any missing ones? Good to know. In that case, I think we can close this? For...

Aaaah right! I remember now. Thanks @stevesloka

After further investigation, I realized that what is actually limiting us is the rate at which the API server can create services. The API server we are using for testing...

One by one. There is no API endpoint that takes a list of services for creation AFAIK.