v1.18¶

Breaking changes

This release contains breaking changes. Review the Breaking changes section below before upgrading. For clusters, follow the rolling upgrade guide.

New features¶

Downlink buffering for idle subscribers: Downlink packets addressed to an idle subscriber are now buffered while Ella Core pages it, and sent to the subscriber once the session is re-established.
Handover data forwarding: Subscribers keep their traffic through 4G and 5G handovers. Ella Core now supports both direct and indirect forwarding, and sends GTP-U End Markers on the old tunnel.
Subscriber registrations: The subscriber API exposes the RAN and core UE identities (RAN/AMF UE NGAP ID, eNB/MME UE S1AP ID). Those can be used for correlating UE events with the radios.
Dark mode: The UI now supports dark mode.
Time range picker: Network events, traffic and audit logs share a common time range picker.
Observability overhaul: Metrics follow Prometheus naming conventions, logs use consistent field names with less noise at info-level, and the eBPF datapath exposes more metrics, including its memory usage in the UI.

Radios are now addressed by their ref field: GET/DELETE /api/v1/ran/radios/{ranNodeType}/{id} -> GET/DELETE /api/v1/ran/radios/{ref} (ex. gNB:001-01:00002a@24)
The deprecated supported TAI field was removed from the radio list endpoint.
On the subscriber detail endpoint, the merged status object is replaced by a registrations array, sessions[].radio_access_type becomes sessions[].system, and ciphering_algorithm and integrity_algorithm report the spec's 128- prefixed names. On the subscriber list, status.radio_access_types becomes status.systems.
Time filters changed format. start and end are now RFC3339 instants. /subscriber-usage records are typed objects ({"date": …} / {"imsi": …}) instead of single-key maps.
Metrics were renamed to follow Prometheus conventions. app_sessions_total -> app_sessions, app_ip_addresses_total -> app_ip_addresses, app_ip_addresses_allocated_total -> app_ip_addresses_allocated, and app_upf_pipeline_latency_nanoseconds_total -> app_upf_pipeline_latency_seconds_total.

If you depend directly on the API, you may need to adapt your integration.