Elastic APM release notes
Review the changes, fixes, and more in each version of Elastic APM.
To check for security updates, go to Security announcements for the Elastic stack.
- Fix Fleet-managed apm-server hanging or crashing on config hot reload due to duplicate monitoring registry (#17872)
- Add config for tail-based sampling discard on write (#13950)
- Add config for tail-based sampling TTL (#16579)
- Truncate string slice attributes in OTLP labels (#434)
- Fix broken UI by explicitly enabling date detection for the
system.process.cpu.start_time
field (#130466) - Use representative count for the
event.success_count
metric if available (#119995) - Fix setting
event.dataset
todata_stream.dataset
ifevent.dataset
is empty, to haveevent.dataset
in everylogs-*
data stream (#129074) - Fix missing trusted root certificate authority in the docker image (#16928)
- Tail-based sampling: Fix missing or infrequent monitoring metric
lsm_size
andvalue_log_size
(#17512) - Fix default tracer
http request sent to https endpoint
error when both self-instrumentation and TLS are enabled (#17293)
- Tail-based sampling: Optimize performance for instances with more CPU and memory (#17254)
- Tail-based sampling: Fix ever-increasing dynamic group count monitoring metric (#17042)
- Tail-based sampling: Fix incorrectly large pebble database
lsm_size
monitoring metric in edge cases which causes TBS to be bypassed (#17149) - Tail-based sampling: Log pubsub errors at error or warn level (#17135)
- Fix missing trusted root certificate authority in the docker image (#16928)
- Tail-based sampling: ignore subscriber position read error and proceed as if file does not exist to avoid crash looping (#16736)
- Tail-based sampling: Storage layer is rewritten to use Pebble database instead of BadgerDB. The new implementation offers a substantial throughput increase while consuming significantly less memory. Disk usage is significantly lower and more stable. See APM Transaction sampling docs for benchmark details. (#15235)