Looking at the ledger for datasets on a sled of dublin, we see (for each disk) a debug, zone, crucible dataset, plus a set of zone/oxz_* for all the zone filesystem roots:
BRM42220026 # cat /pool/int/2196ba5d-e211-4495-9c0b-e0948ba0afbd/config/omicron-datasets.json | jq '.datasets' | grep kind | sort
"kind": "cockroachdb"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "external_dns"
"kind": "internal_dns"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone/oxz_cockroachdb_d5dfdf09-a943-4e8c-a7f0-18e1e1644dae"
"kind": "zone/oxz_crucible_27e2564b-4d14-4ba7-a198-7fd39c0345b4"
"kind": "zone/oxz_crucible_36952839-104d-418c-a6cb-cc509b12531e"
"kind": "zone/oxz_crucible_4586ae39-ec1c-4d1e-a36c-4c1e6621ec3c"
"kind": "zone/oxz_crucible_583f9a41-2587-436e-8b1d-c31892ec84ff"
"kind": "zone/oxz_crucible_722a3bd6-09d1-49af-8001-aabd2d8fd2c5"
"kind": "zone/oxz_crucible_75936029-f293-40fb-96d5-f434332953cc"
"kind": "zone/oxz_crucible_e5a997e7-1551-4d9f-95b1-ede53409d0a0"
"kind": "zone/oxz_crucible_e8b9f2da-79c4-4ded-a7e4-631f55429c2a"
"kind": "zone/oxz_crucible_f3c60b6c-76cf-4535-b07b-9fd97abca79c"
"kind": "zone/oxz_crucible_pantry_462c1eae-4f59-4314-8308-fea70aa6ba5d"
"kind": "zone/oxz_external_dns_7bc405cf-c950-48ef-975e-d80756d67292"
"kind": "zone/oxz_internal_dns_2fdd5688-482b-40d7-b16b-b2c01715e6fc"
"kind": "zone/oxz_ntp_68ec5db8-1f62-4640-ba51-44c09cba28de"
"kind": "zone/oxz_oximeter_a97b1d79-18a2-4589-b7c3-a9bb3929e254"
However, on a sled of dogfood, none of the zone/oxz_* filesystem roots are present; only the durable datasets:
BRM42220014 # cat /pool/int/7c377f39-95bf-4074-8dda-7a7a102b9d2c/config/omicron-datasets.json | jq '.datasets' | grep kind | sort
"kind": "cockroachdb"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "crucible"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "debug"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
"kind": "zone"
This appears consistent with the blueprint: the current target blueprint has filesystem_pool: null for all zones. However, it is surprising; we've been doing work under the assumption that this value is set for all zones, and disk expungement may not work correctly (expunging a disk should expunge any zones that depended on that disk, but that assumes reconfigurator knows what zpools the zones are using!).
Looking at the ledger for datasets on a sled of
dublin, we see (for each disk) a debug, zone, crucible dataset, plus a set ofzone/oxz_*for all the zone filesystem roots:However, on a sled of dogfood, none of the
zone/oxz_*filesystem roots are present; only the durable datasets:This appears consistent with the blueprint: the current target blueprint has
filesystem_pool: nullfor all zones. However, it is surprising; we've been doing work under the assumption that this value is set for all zones, and disk expungement may not work correctly (expunging a disk should expunge any zones that depended on that disk, but that assumes reconfigurator knows what zpools the zones are using!).