Skip to content

2026.07.14.1530

@konrad-ohms konrad-ohms tagged this 14 Jul 15:30
This agent release updates the following bundles:

* agent 1.1.781: Adds runtime configuration override delivery per plugin and process, adds a non-root agent support task, enables host ID persistence by default for v2 assemblies, and caps NO_CHANGE reporting at 60 seconds for Infra Smart Alerts compatibility.
* agent-ibmi-handler 1.0.5: Fixes IBM i JVM job name resolution for systems with a large number of active JVMs.
* agent-jvm-handling 1.0.52: Makes JVM attach and load concurrency configurable via environment variables.
* agent-process-handling 1.0.53: Fixes zombie and dead process detection on z/OS and IBM i by enriching bulk process scans with process state information.

Features

This release adds the following features:

* The backend can now push runtime configuration overrides into the agent's plugin configuration delivery layer per plugin name and process ID without modifying any on-disk YAML files. Overrides are deep-merged at delivery time and are scoped per (pluginName, pid) pair.
* A new support task is available for non-root agent deployments. When collecting support information, it reports the agent's Linux capabilities, key path access permissions, and the instana-sudo / JVM attach strategy to help identify permission issues.
* For v2 agent assemblies, host ID persistence is now enabled by default. The agent detects the assembly version from the VERSION file and sets INSTANA_PERSIST_HOST_UNIQUE_ID=true automatically. v1 behaviour is unchanged; set INSTANA_PERSIST_HOST_UNIQUE_ID=false to opt out on v2.
* JVM attach and load concurrency limits are now configurable via INSTANA_JVM_CONCURRENT_ATTACHMENTS (default: 4) and INSTANA_JVM_CONCURRENT_LOADS (default: 6).

Fixes

This release includes the following fixes:

* Previously, on IBM i systems with many active JVMs, the database call used to resolve JVM job names was slow and frequently timed out, causing JVMs to appear with meaningless 4-digit names in the UI with no retry. Now, job names are resolved via a process call instead, making resolution reliable regardless of the number of active JVMs.
* Previously, zombie and dead processes were not correctly excluded on platforms using the native Sigar library (including AIX, Solaris, and HP-UX) because process state was not retrieved. On z/OS and IBM i, state was also not captured during bulk process scans. Now, process state is queried via Sigar on all supported platforms, and on z/OS and IBM i it is additionally parsed from enriched ps output, ensuring zombie and dead processes are consistently identified and excluded.
* Previously, the NO_CHANGE signal could be sent at intervals longer than 60 seconds for sensors with high poll rates, causing gaps in Infra Smart Alerts data. Now, NO_CHANGE is sent at most every 60 seconds regardless of the configured poll rate.

Internal updates

These changes improve internal stability and prepare future enhancements:

* agent-runtime 1.0.23
Assets 2
Loading