https://pagure.io/centos-infra/issue/1528

We have for some years 4 supermicro chassis that were used for various workloads, including CI.
These chassis have blade that are actually unused (previously on premises Openshift OCP cluster, or else) and some that are still in use for prod workload.
As all these chassis are old and not covered by any warranty status, we'll have to shuffle services around and also at least remove 2 chassis from racks.

Proposal :

• identify workload
• see what is "critical" and what's not
• reduce footprint with only keeping two unwarrantied chassis
• for important/critical workload, just prepare HA plan to quickly move from one chassis to remaining one
• for the rest (like koji builders, etc), just split load across two chassis
• retire/decom two unused chassis