This guide covers what changes when upgrading from v1.6 to v1.7. There is one breaking change and four behavior changes to review before upgrading.

There are no new PostgreSQL or ClickHouse migrations. Upgrade by replacing the image in place, with no extra steps. Retries scheduled by v1.6 run normally on v1.7, and you can roll back to v1.6.0 the same way.

Change	Impact	Action Required
Redis and message queue OpenTelemetry instrumentation is off by default	Deployments that export OpenTelemetry traces or metrics	Set OTEL_ENABLED_INSTRUMENTATIONS=redis,mq to keep them

For new features and fixes, see the v1.7.0 release notes.

Breaking Changes

Redis and message queue OpenTelemetry instrumentation is off by default

With OpenTelemetry enabled, Outpost no longer exports Redis command spans or internal message queue client spans and metrics. Outpost's own traces and metrics, and the HTTP server spans, are unchanged. Without OTEL_SERVICE_NAME set, nothing changes.

To export them again, set OTEL_ENABLED_INSTRUMENTATIONS=redis,mq, or only one of the two values, for example OTEL_ENABLED_INSTRUMENTATIONS=redis. Any other value fails startup. See Observability.

The GCP Pub/Sub destination's gRPC spans and metrics are no longer exported either, and no setting turns them back on.

Behavior Changes

• Retry delays shorter than LOG_BATCH_THRESHOLD_SECONDS (default 10) are raised to it. This affects RETRY_SCHEDULE steps and a RETRY_INTERVAL_SECONDS below it; the default retry timing (RETRY_INTERVAL_SECONDS=30) is unchanged. Outpost logs a warning at startup that lists the affected delays. See Automatic retries.
• At most RETRY_MAX_CONCURRENCY retries (default 100) start at once per process. v1.6 had no limit; set it to 0 to keep it that way.
• A tenant JWT whose deployment_id claim doesn't match DEPLOYMENT_ID is rejected with 401. Tokens issued by Outpost and tokens without the claim keep working.
• A path or query parameter containing a NUL character or invalid UTF-8 returns 400, where the PostgreSQL log store returned 500.

New Configuration

See the configuration reference for defaults and details.

• RETRY_MAX_CONCURRENCY: limit concurrent retry starts.
• RETRY_MAX_RECEIVE_COUNT: failed retry starts before the retry moves to the dead-letter queue. v1.6 fixed this at 5, which stays the default.
• RETRY_LOOKUP_BATCH_SIZE and RETRY_LOOKUP_BATCH_WAIT_MS: combine retry lookups into fewer log store queries. Off by default.
• CLICKHOUSE_MAX_OPEN_CONNS, CLICKHOUSE_DIAL_TIMEOUT_SECONDS and CLICKHOUSE_READ_TIMEOUT_SECONDS: ClickHouse connection pool size and timeouts. The defaults match v1.6.
• OTEL_ENABLED_INSTRUMENTATIONS: opt in to Redis and message queue instrumentation.
• DESTINATIONS_WEBHOOK_TIMESTAMP_FORMAT accepts iso8601_milli_no_offset.