Datacake - Upgrade The Things Stack Cloud to v3.36.2 – Maintenance details

During this maintenance window we will upgrade The Things Stack Cloud to v3.36.2.

We expect brief service interruptions during this deployment.

Some gateways may temporarily disconnect as services are restarted. Gateways are expected to automatically reconnect, but in some cases this may take up to one hour.

Here is the changelog since the current version v3.36.2:

Added

• Redis connection buffer sizes — read-buffer-size and write-buffer-size on any Redis config (e.g. as.redis.read-buffer-size). Default 0 = 32 KiB per buffer. Lower these on deployments with many connections (such as the Application Server's per-subscription Pub/Sub connections) to cut memory usage.
• MONITORING_STOPPED alert status — for ALERT_GATEWAY_DISCONNECTED, the Alert Routing Server now checks the gateway's live state with the Gateway Server on resolve, and reports MONITORING_STOPPED when monitoring ended while the gateway was still offline, instead of a RESOLVED that reads as "gateway back online".
• Direct LNS for Semtech RJS claims — dcs.gcls.semtech-rjs.lns-uri and dcs.gcls.semtech-rjs.lns-trust. When set, claimed gateways are pointed at this LNS instead of the CUPS. Fixes claiming of The Things Indoor Gateways.
• Event stream allowlist — events.redis.store.event-type-names limits which events are written to the per-event-type streams that group subscribers (e.g. the Network Operations Center) read. Empty by default, which preserves current behavior. Setting it to only the consumed events reduces write load on the events Redis.

Changed

• Gateway disconnection alerts may now deliver the MONITORING_STOPPED status to webhook, email and SMS receivers.
• Per-entity event stream TTLs are refreshed at most once per half of the configured history TTL instead of on every event, lowering events Redis CPU. Retention is unchanged.
• The end device Get rate limit (Identity, Network, Application and Join Server registries) is now per application instead of per device. Clients reading many devices in one application share a single limit.

Fixed

• Identity Server user and organization reads no longer scan the whole accounts table — lookups are now tenant-scoped so the accounts (tenant_id, uid) index is used.
• MAC settings profile values are now applied to the device's MAC state during join (OTAA) and factory reset (ABP).
• Federated (OIDC/SSO) login no longer fails with linked_user (linked user not found) after a linked local user is purged. Purging a user now also removes its external-user associations.
• Claiming a The Things Indoor Gateway with an invalid owner token returns an invalid argument error (HTTP 400) instead of a 401, so the Console shows a claiming error rather than a "Please sign in again" dialog.
• The as.webhook.fail event again carries the upstream message's correlation IDs when webhook retries are enabled, so FindRelatedEvents can trace failed deliveries.