Upgrading to 0.2.0
Swap the binary or image on the same bucket; what Cello reads, writes, and no longer speaks.
Cello 0.2.0 keeps your data and changes the wire. Stop every node of the
old fleet, then start Cello 0.2.0 on the same bucket: celld-written data
carries over, new writes use the cello spellings, and a mixed fleet of
old and new nodes is unsupported.
Upgrade steps
- Stop all old nodes. Cello 0.2.0 speaks a new peer protocol
(
cello-tunnel,x-cello-overload,cello-log-stream, and a new overload marker), so an old node and a new node cannot share a fleet. There is no rolling upgrade from denoland/celld or Cello 0.1.x: stop every node first, then start 0.2.0. - Swap the binary or image.
ghcr.io/ryuzcorp/cello:<semver>(0.2.0,0.2,beta,latest) replacesghcr.io/ryuzcorp/celld:<version>-ryuzcorp.<n>. Consumers pin by digest as before. The bucket, its prefixes, and the node data directories stay where they are. - Rename environment variables. Every
CELLD_*variable becomesCELLO_*, strictly one-to-one, with no fallbacks and no aliases:CELLD_BUCKET→CELLO_BUCKET,CELLD_OPERATOR_TOKEN→CELLO_OPERATOR_TOKEN,CELLD_DEV_STATE→CELLO_DEV_STATE, and so on. The node,cello dev, and every CLI refuse to start when anyCELLD_*variable is set, naming theCELLO_*spelling — a carried over environment can never silently lose config. In particular a staleCELLD_OPERATOR_TOKENnever leaves the internal listener open: the node fails closed before serving anything. - Start Cello on the same bucket. No export, no import, no rewrite pass. Old data stays as it is until it is naturally rewritten.
What carries over
Persisted identifiers are written with the cello spelling and read
with both, in one place per family, so the old spelling is obvious and
easy to delete later:
- The runtime isolate pool script: new graphs record
__cello-runtime; graphs that record__celld-runtimestill resolve to the runtime pool. - Stored stub rows: rows written before the binding-isolation patches still decode as plain data on first read; storing them again writes the new envelope.
- The R2 user-metadata envelope: new objects carry
cello-r2(cello_r2on Azure Blob Storage); objects carryingcelld-r2read the same way. Cursors work under both thecello-r2-cursor-v1.and thecelld-r2-cursor-v1.spellings; new cursors use the new one. - Telemetry batches: the retention sweep lists the date-partitioned key
and never reads the forensic stamp, so batches stamped
celld-schema/celld-retentionkeep their lifecycle. New batches are stampedcello-schema/cello-retention. Readers of the stamp itself must accept both spellings. - Container image references: manifests that record
celld-image:<key>resolve to the samedeploy/images/<key>.tarascello-image:<key>; new manifests record the new spelling.
Upstream denoland/celld#NNN links in comments and docs are historical
references and still resolve.
What does not carry over
- The wire. The subprotocol, the overload signal, the log-stream upgrade value, and the overload marker all use new spellings. Mixed fleets are unsupported: a 0.2.0 node never peers with an older one.
- In-process names. The isolate↔host
__celloglobal, the__cello*tokens, the__CELLO_*__sentinels, and the test-only markers are never persisted; they were renamed outright. - Local state.
cello devkeeps its state in.cello/dev; the old.celld/devdirectory is not read. - Health and telemetry names moved from
celld.*tocello.*in 0.1.0 already (cello.versionreports the build); update dashboards and liveness probes if you have not.