done.

https://fault-tolerance.toolforge.org/map?cluster=ms+backend

ladsgroup merged https://gitlab.wikimedia.org/repos/data_persistence/fault-tolerance/-/merge_requests/7

We need to update the records of hosts manually from time to time. The plan is to actually productionize the service soonTM so we can leave the netbox token in the code and it would look it up automatically but since then, I need to update the record once every couple of weeks.

Yes, for my purposes, considering e.g. thanos-be* as one cluster would be entirely fine. Sorry, I meant to reply to your previous.

Would that work for you?

In T389612#10661875, @MatthewVernon wrote:

My other dumb idea would be: netbox knows the answer, and has http endpoints, could that be the solution?

My other dumb idea would be: netbox knows the answer, and has http endpoints, could that be the solution?

I'm sorry, I don't think I do. The swift (ms and thanos) rings are on the puppetserver and thus available (but probably not by https), but you don't really want to be guddling through them. And the internals of the Ceph clusters aren't exposed to the outside world much because nothing needs to interact with them except the clients (which are the S3 gateways) :-/

Thanks :)

Reading from puppet repo is not super nice, it's possible but it's brittle, do you have an http endpoint I can reuse?

Updated netbox file to the latest and it now shows up. Once this is productionized, it would be done automatically.

They're in puppet - for swift it's profile::swift::storagehosts: (different per DC), for thanos it's profile::thanos::swift::backends:, and for apus it's in cephadm_clusters in hieradata/{codfw,eqiad}.yaml

Where can I find the list of them? For example, db sections can be found via https://noc.wikimedia.org/dbconfig/eqiad.json and front-end are being discovered via https://config-master.wikimedia.org/pools.json (powered via pybal I think)