As per title, all hosts running Graphite should be on Bullseye
```
root@cumin1001:~# cumin 'P{C:graphite} and not P{F:lsbdistcodename = buster}'
2 hosts will be targeted:
graphite2003.codfw.wmnet,graphite1004.eqiad.wmnet
DRY-RUN mode enabled, aborting
```
[x] Get the `graphite::production` role to work in Pontoon on Bullseye
Action plan for codfw:
[x] Make sure `carbonate` is available on Bullseye and working as expected, as per https://wikitech.wikimedia.org/wiki/Graphite#Merge_and_sync_metrics
[x] Reimage graphite2003 with Bullseye. graphite1004 will start dropping metrics directed to graphite2003 as expected
[x] Ensure Puppet is running as expected, all services are up and metrics are received from graphite1004 and wait 24h for some data to accumulate
[x] Transfer and merge data from graphite1004 to graphite2003 with carbonate, following https://wikitech.wikimedia.org/wiki/Graphite#Merge_and_sync_metrics
[x] Validate that historical metric data is present and new data is flowing
The plan for eqiad is similar, with the addition of a failover to codfw as per https://wikitech.wikimedia.org/wiki/Graphite#Failover and fail back once things are working in eqiad.
[x] Failover out of graphite1004 and to graphite2003
[x] Reimage graphite1004
[ ] Transfer and merge data from graphite1004 to graphite2003 with carbonate, following https://wikitech.wikimedia.org/wiki/Graphite#Merge_and_sync_metrics
[ ] Validate that historical metric data is present and new data is flowing
[ ] Failover back to graphite1004