Release 21.2.15

Updated November 13, 2026 17:15

Article Published Date: 11/13/2026

ZEDEDA Cloud versions

This patch release contains fixes only. No new features or enhancements are included.

Review Changed Behavior before updating dashboards or alerts on OpenTelemetry metrics.

Users with project-scoped roles no longer receive a 403 Forbidden error when enabling Edge View or reading the configuration of an edge node in an authorized project.

Edge nodes running EVE-OS 13.4 LTS no longer re-download application content they already had after an unrelated configuration change. Content URLs remain the same unless the content itself changes.

Applies only to enterprises with an OpenTelemetry (OTEL) data stream.

Memory values of 4 GiB or greater are now reported correctly.
Memory and disk metrics now use the unit bytes. They were incorrectly labeled as seconds.
Running totals are now exported as cumulative monotonic counters (OTLP Sum) instead of gauges. This covers CPU time, network traffic, network errors and drops, Edge Infrastructure Services API call counts, ZFS pool and volume I/O, and patch envelope counters.
When an application instance has several network interfaces, each interface now carries a distinct if_name attribute. The same correction applies to volume_uuid on ZFS volume metrics.
ZFS pool metrics now include a pool_name attribute.
Patch envelope metrics are now one series per application instance, summed across its envelopes.
device_cpu_system_total has been removed. It reported 0 for all edge nodes.
Metric descriptions are added and corrected. For example, drop counters count packets, not bytes.

Review the following changes before updating dashboards or alerts that query OpenTelemetry metrics.

Area	Before 21.2.15	From 21.2.15	What to do
OpenTelemetry metrics — units	Memory and disk metrics were labeled `seconds`.	Memory and disk metrics use `bytes`.	Update dashboards and alerts that query memory or disk metrics.
OpenTelemetry metrics — counter types	Running totals (CPU time, network traffic, network errors and drops, API call counts, ZFS I/O, and patch envelope counters) were exported as gauges.	Running totals are exported as cumulative monotonic counters (OTLP Sum).	Update queries or alert rules that reference these metrics.
OpenTelemetry metrics — interface and volume attributes	All network interfaces on a multi-interface application instance shared the same `if_name`. All ZFS volume metrics shared the same `volume_uuid`.	Each interface and each ZFS volume carries a unique attribute value.	Update dashboards or alerts that filter by `if_name` or `volume_uuid`.
OpenTelemetry metrics — removed metric	`device_cpu_system_total` was always 0.	`device_cpu_system_total` is removed.	Remove any queries or alerts that reference `device_cpu_system_total`.

Was this article helpful?

0 out of 0 found this helpful