Travis Nielsen

@rzarzynski Any insights to this osd call stack?

> @travisn done Thanks, though looks like the same error is happening for the pool resize test...

@Hello-Linux Try accessing the dashboard by creating another service instead of using the internal cluster ip. See the [external access dashboard topic](https://rook.io/docs/rook/latest/Storage-Configuration/Monitoring/ceph-dashboard/#viewing-the-dashboard-external-to-the-cluster).

Looks like the address to the nodeport should be https://10.68.0.166:30770. The browser is updating that to 10.68.110.110:8443? I don't see in your example where the other endpoints come from. Does...

Does the mgr pod log show the dashboard request even when the endpoint is rewritten? If the request is not in the log, there must be some routing issue outside...

Ceph must have that limit because it's really not designed to run with hundreds of pools and so many PGs. Are you creating so many object stores because you need...

> > Also if the use case is to have multiple ObjectStore daemons (RGWs) to provide more client connections and more parallel work behind LB that is fully supported in...

Multiple rgw pods can run on a node, it's not a problem unless host networking is configured. Each object store could configure where its rgw pods run with the [placement](https://github.com/rook/rook/blob/release-1.10/deploy/examples/object.yaml#L58-L81),...

> Maybe we can somehow create different Pools using one Crush rule? Even if different pools are created with the same Crush rule, Ceph will quickly hit other limits, such...

> Is it impossible to create a bucket claim with the help of S3 API? Our goal is to create many independent repositories controlled via S3 API. Are you saying...