Upgrading to 2.0

Version 2.0 introduces the riot distribution alongside riotx (see Installation). Identifiers that used the riotx name now use riot; only the riotx distribution and its executable keep the riotx name.

Breaking changes

Docker images

Images are now published as redis/riotx and redis/riot. The previous riotx/riotx image stays at RIOT-X 1.15.x and riotx/riot stays at RIOT 4.3.0; neither receives new releases. Update Docker commands, Kubernetes manifests and Helm values when upgrading. For the collector chart, change global.image.repository (or an explicit riotx.image.repository override) from riotx to redis. Container launchers remain available on PATH and at /riotx/bin/riotx or /riot/bin/riot.

Commands

sync2 is renamed sync, and replicate no longer accepts sync as an alias. Scripts that ran riotx sync (meaning replicate) or riotx sync2 must be updated.

riot distribution

riot now ships only the reactive, Spring-free commands: sync, pubsub, diff (alias compare), gen, stats and ping. Their options differ from the riotx commands of the same name (except sync and pubsub, which are identical). Use riotx for replicate, faker, imports and exports, and JavaScript expressions (--key-proc, --ttl-proc, --channel-proc).

Replication metrics

Metrics are renamed from riotx.replication. to riot.replication.. In Prometheus this means riotx_replication_bytes_total, riotx_replication_lag_seconds and riotx_replication_read_latency_seconds become riot_replication_bytes_total, riot_replication_lag_seconds and riot_replication_read_latency_seconds. Update dashboards and alerts accordingly; the Grafana dashboard shipped with RIOT already uses the new names. The default metrics name changes from RIOTX-<version> to RIOT-<version>.

Stream consumer group

The default consumer group for stream-import changes from riotx-<version> to riot-<version>. The default already changed with every version, so pass --consumer-group to keep reading with an existing group.

Change event source name

The default source name in change events changes from RIOT-X to RIOT.

Command-line option changes

Sync options in both distributions

riot sync and riotx sync now use the same canonical option names. The former names are rejected rather than retained as hidden aliases.

Former option Replacement

--no-replace

--ignore-existing

--key-include, --key-exclude

--include, --exclude

--diff, --compare

--verify

--remove-source-keys

--delete-source

-n is no longer accepted by either sync command. Use --dry-run to preview copies without writing data, or --ignore-existing to skip existing target keys. Single-server -n / --db is unchanged.

Thread and stream options also use shorter names, such as --source-threads, --target-threads, --stream-poll, --stream-reconcile, and --max-streams. See the complete sync option migration table, including the connection options shared with other reactive commands. For single-server reactive commands, --computation-threads becomes --threads.

Sync verification defaults to RIOT_VERIFY, then falls back to RIOT_DIFF, then RIOT_COMPARE, then QUICK. An explicit --verify overrides all three environment variables. The established replicate command retains its existing sync-related options, including --diff.

TLS verification options across commands

TLS option names change in both distributions, including riotx replicate, imports, and exports. Update scripts using any of these former names:

Former TLS option Replacement

--verify

--tls-verify

--source-verify

--source-tls-verify

--target-verify

--target-tls-verify

The former TLS names are rejected. sync --verify controls data verification after copying; it does not configure TLS. The environment variables RIOT_REDIS_TLS_VERIFY, RIOT_SOURCE_TLS_VERIFY, and RIOT_TARGET_TLS_VERIFY keep their existing names. See TLS verification modes for examples and accepted values.

Reader limit changes in riotx

Import and generation commands now share --count / RIOT_COUNT, applied to each outer reader before processing and filtering. Imports default to unlimited; faker and gen default to 1000 and require a positive count. Generator --count 0 and RIOT_COUNT=0 are rejected; do not use zero as a generator dry run.

Previous setting

Replacement

--max N

--count N (no compatibility alias)

RIOT_MAX=N

RIOT_COUNT=N; RIOT_MAX is no longer read

JDBC --rows N

--max-rows N; the environment variable remains RIOT_ROWS

stream-import --count N (XREAD batch size)

stream-import --read-count N

stream-import --block DURATION

stream-import --read-block DURATION

On stream-import, --count N now stops after N source messages across the configured streams. An old command such as stream-import s1 --count 50 hset …​ therefore stops after 50 messages instead of requesting batches of 50 indefinitely. Change it to stream-import s1 --read-count 50 hset …​ to preserve the old behavior.

RIOT_COUNT now affects database, file, Redis, Stream, and Elasticsearch/OpenSearch imports as well as CDC imports and generators. If it was exported for generation, unset it before an unlimited import or explicitly pass --count 0. Replace deployed RIOT_MAX values with RIOT_COUNT; leaving only RIOT_MAX no longer limits the import.

Snowflake and Databricks count rows per table across snapshot and CDC batches, including --gen. They reject positive JDBC --max-rows / RIOT_ROWS limits before opening connections: truncating a batch result can skip unread changes. Use the outer --count limit instead. db-import still supports --max-rows for JDBC Statement.setMaxRows().

Snowflake JDBC pool sizing

snowflake-import now sizes its JDBC pool to max(10, number of tables + 2) when --jdbc-pool and RIOT_JDBC_POOL are unset. Imports of nine or more tables can therefore open up to number of tables + 2 Snowflake connections instead of ten. An explicit --jdbc-pool is used as-is; a value below the number of tables logs a warning at startup. No configuration change is required.

Migrated automatically

CDC offsets

The default offset key prefix changes from riotx:offset: to riot:offset:, and the default offset hash from metadata:riotx:offsets to metadata:riot:offsets. When the default is used and no offset exists under the new name, the offset stored under the old name is read, so CDC imports resume where they left off. New offsets are written under the new name. --offset-clear removes both. Explicitly configured --offset-prefix or --offset-key values are used as-is.