Release notes for Instana agent bundles

This release adds the following features:

• The agent UI now discovers and displays logs from custom log paths that are configured in org.ops4j.pax.logging.cfg. This change improves visibility into application logs stored in nonstandard locations.
• The IBM i sensor can now access AS400 objects directly from AS400Connection instances, enabling more efficient integration with IBM i systems.

This release includes the following fixes to improve reliability and correctness:

• Previously, the agent incorrectly deleted hsperfdata files for Java processes in the ignore list and generated unnecessary cleanup messages. Now, the agent only stops monitoring these processes without deleting their attach files.
• Previously, the agent misidentified transitioning processes as JVMs and send SIGQUIT signals to non-Java processes, which caused them to quit unexpectedly. Now, you can configure the agent to attach only to JVMs with valid attach files by using INSTANA_JVM_HANDLING_REQUIRE_ATTACH_FILES=true. This option is disabled by default to maintain existing attachment behavior.
• Previously, Netty version 4.1.125.Final contained the security vulnerability CVE-2025-67735. Now, Netty has been updated to version 4.1.129.Final to address this vulnerability.

These updates contain internal improvements that do not change customer-facing behaviour but ensure long-term stability and readiness for future enhancements:

• thirdparty-netty 4.1.129.Final
• agent-process-handling 1.0.37
• sensor-api 1.2.177

This release adds the following features:

For more information, see Legacy OTLP gRPC port .

This release includes the following fixes to improve reliability and correctness:

• Previously, process start times changed during agent restarts, which triggered false service alerts. Now, process start times remain stable across agent updates and restarts.
• Previously, deleting javaagent-loader JAR file at runtime caused NoClassDefFoundError failures. Now, the agent pins all loader classes at startup to prevent class-loading errors.
• Previously, the agent incorrectly excluded JVMs with custom class loaders such as WebLogic's LaunchClassLoader from attachment. Now, the agent excludes only JVMs with unsupported SAS custom class loaders.
• Previously, the agent exposed environment variables that contained secrets in Containerd container information when you use the CRI client. Now, the agent masks secrets properly in all container configuration locations.

These updates contain internal improvements that do not change customer-facing behaviour but ensure long-term stability and readiness for future enhancements:

• InstanaPCP 1.3.79
• agent 1.1.763
• agent-runtime 1.0.17
• agent-runtime-api 1.0.14

This release adds the following features:

• The agent now validates OSGi bundle start-levels when creating sensor configurations and provides actionable warnings when sensors fail to start due to mode mismatches. These warnings help operators quickly identify and resolve configuration issues.
• Backend connectivity errors are now logged with throttled warnings for HTTP 402 (Payment Required) and HTTP 404 (Not Found) responses, providing better visibility into endpoint access restrictions and configuration issues.
• JVMs running with custom class loaders are now automatically detected and reported through monitoring events in the agent dashboard, providing clear visibility into why dynamic attachment is not supported for these processes.

This release includes the following fixes to improve reliability and correctness:

• Previously, sensor payloads from multiple sources (such as legacy and sidecar sensors for IBM MQ) might overwrite each other when arriving simultaneously, causing data inconsistencies. Now, the agent performs deep merging of JSON payloads to preserve all data from both sources.
• Previously, JVM attachment might fail in Podman containers when the /tmp directory was mounted from the host root file system, causing permission errors with OpenJ9 attach files. Now, the agent automatically retries the attachment without user switching when the initial attempt fails, ensuring successful monitoring in these environments.
• Previously, on Windows systems, orphaned InstanaPCP.exe processes might remain to run after agent crashes or forced terminations, preventing new instances from starting on agent restart. Now, the agent detects and terminates leftover processes automatically, restoring Windows monitoring capabilities.
• Previously, InstanaPCP metrics collection failed after transitioning to flexible process name handling due to inconsistent instance name formats. Now, the agent uses consistent naming formats for all lookups, restoring accurate metrics collection.
• Previously, the pollRate configuration was only sent after the first reset (10 minutes after sensor start), causing gaps in UI graphs during the initial period. Now, pollRate is sent immediately on the first sense and then on subsequent resets, ensuring continuous graph visualization.
• Previously, InstanaPCP processes on Windows did not shut down properly when the Instana Agent Service stopped, leading to multiple parallel instances on restart. Now, InstanaPCP correctly monitors the agent endpoint and terminates when the agent stops, preventing duplicate processes.

This release includes the following fixes to improve reliability and correctness:

• Previously, CRI-API based process handling (used with CRI-O and containerd) might hang or leave background threads active when command output was not fully consumed. This behavior might affect long-term stability. Now, background tasks and I/O resources are properly cleaned up, which can ensure stable and reliable process monitoring for CRI-O and containerd environments.

These updates contain internal improvements that do not change customer-facing behavior but ensure long-term stability and readiness for future enhancements:

• agent-util 1.0.16

This release adds the following features:

• OpenTelemetry base components are now always started by default, even when OpenTelemetry is not explicitly configured, providing better readiness for telemetry collection.
• Podman containers can now be monitored even when no Podman socket is available or when the socket is reported as missing, enabling monitoring in more restrictive environments.
• Message size limits can now be configured for the Agent endpoint (:42699) by using INSTANA_AGENT_HTTP_MAX_MESSAGE_SIZE (minimum: 5 MB, maximum: 49.5 MB), allowing better control over resource usage.
• gRPC message size limits can now be configured by using INSTANA_AGENT_OTEL_GRPC_MAX_MESSAGE_SIZE or through config key grpc.max_message_size (minimum: 5 MB, maximum: 49.5 MB), providing flexibility for large telemetry payloads.
• HTTP message size limits can now be configured by using INSTANA_AGENT_OTEL_HTTP_MAX_MESSAGE_SIZE or through config key http.max_message_size (minimum: 5 MB, maximum: 49.5 MB), enabling handling of larger HTTP-based telemetry data.
• Fine-grained control is now available for running processes within the context of other processes (for example, inside containers). You can now control how environment variables are inherited and how user context is applied.
• The Agent Support Task now includes Podman socket detection details in its downloadable results, improving troubleshooting capabilities.
• The Agent now validates potential gateway endpoints inside the Podman process network namespace before applying them, ensuring more reliable network configuration.

This release includes the following fixes to improve reliability and correctness:

• Previously, evaluation of performance counters and temporary directories was incorrectly bound to the InstanaPCP state (introduced in 1.0.27), causing these checks to fail when InstanaPCP was not active. Now, temp directories are always evaluated, and performance counters are queried on Windows regardless of InstanaPCP state, ensuring consistent system monitoring.
• Previously, additional information requests for user-context containers in rootless Podman monitoring might fail due to shell output contamination. Now, the Agent uses a user session without a login shell to ensure clean stdout output, enabling reliable data collection from rootless Podman containers.
• Previously, environment variables during java -version checks and JVM attachment might interfere with proper operation. Now, a cleaned-up environment is used for these operations, ensuring more reliable JVM detection and attachment.

These updates contain internal improvements that do not change customer-facing behavior but ensure long-term stability and readiness for future enhancements:

• instana-agent-open-telemetry-base 2.2.5
• instana-agent-otel-sensor-sdk 1.1.18
• instana-agent-open-telemetry-service 1.1.13
• agent-util 1.0.15
• agent-jvm-handling 1.0.29
• agent 1.1.761
• agent-backend 1.0.28
• agent-backend-api 1.0.13
• agent-backend-util 1.0.13
• agent-configuration-integration-vault 1.0.11
• agent-grpc 1.0.13
• agent-grpc-channel-cache 1.0.13
• agent-http 1.0.13
• agent-jaeger-tracer 1.0.13
• agent-memory 1.0.13
• agent-netty-allocator 1.0.13
• agent-netty-allocator-api 1.0.13
• agent-process-handling-api 1.0.13
• agent-self-profiling 1.0.13
• agent-sensor 1.0.13
• agent-socket-server-api 1.0.13
• agent-task-orchestrator 1.0.13
• agent-zipkin-tracer 1.0.13

This release includes the following fixes to improve reliability and correctness:

• Previously, sensors with pollRate 1 had a 10-second gap before heartbeats were converted to NO_CHANGE results, even though they were expected to provide data every second. This delay caused special heartbeat workarounds to malfunction when sensors lagged. Now, the conversion threshold for pollRate 1 sensors is reduced to 3 seconds, which ensures more timely data processing and correct handling of lagging sensors.
• Previously, container ID detection could produce false positives in mixed container runtime environments, particularly when CRI-O containers were incorrectly identified by Containerd detection logic. Now, container detection uses explicit validation checks and runtime-specific patterns (including improved Podman libpod scope detection), ensuring accurate container identification across Docker, Containerd, Podman, and CRI-O environments.
• Previously, the agent could not properly interact with processes inside Podman containers, especially rootless ones, due to user namespace differences. This issues caused permission errors and failed operations when attaching to JVMs inside these containers. Now, the agent intelligently detects namespace relationships and uses appropriate mechanisms to join target namespaces, ensuring reliable JVM attachment and monitoring in Podman environments.

These updates contain internal improvements that do not change customer-facing behavior but ensure long-term stability and readiness for future enhancements:

• agent-internal-api 1.0.18
• agent-jvm-handling 1.0.28
• agent-runtime 1.0.16
• agent-runtime-api 1.0.13

This release includes the following fixes to improve reliability and correctness:

• Previously, container ID detection for Podman POD containers was incomplete. Now, detection is enhanced to properly identify POD containers, improving container visibility and monitoring accuracy.

These updates contain internal improvements that do not change customer-facing behavior but ensure long-term stability and readiness for future enhancements:

• sensor-api 1.2.170

This release adds the following features:

• The agent bundle introduces a new configuration option to control how configuration file changes are detected. You can enable with the environment variable INSTANA_WATCH_CONFIG_CHANGES_WITH_HASH_POLLER (default: false; set to true to use hash-based polling instead of inotify).
• The agent-runtime and agent-runtime-api bundles improve Linux process detection:
  • The runtime now detects if /proc is present in Linux paths.
  • The API now exposes this information through the isProcPresent method.

• Previously, the process abstraction consumed more memory when it was checking for /proc presence. This issue is now fixed by optimizing these checks.
• Previously, process startup time detection on AIX was less accurate. This issue is now fixed by using SIGAR for the detection.

This release adds the following features:

• The agent-init bundle enables direct memory usage on AIX systems by default. This can be turned off with the environment variable INSTANA_AIX_NETTY_SET_UNSAFE (default: false; set to true to disable direct memory).

• Previously, the AgentSupportTask that executes ps might omit the parent process ID (PPID) in some cases. This issue is now fixed by probing the available ps options on startup and selecting one that reliably includes the PPID and, when possible, the process tree.
• Previously, the agent-jvm-handling bundle might not attach to JVMs running inside Podman containers. This issue is now fixed by using switchUser during the attached operation.
• Previously, the agent-netty-allocator bundle lacked visibility when direct memory was disabled. This issue is now fixed by adding an information-level log to indicate the state.
• Previously, the agent-process-handling bundle did not correctly apply switchUser in Podman processes that rely on nsenter. This issue is now fixed by implementing switchUser with the --setuid and --setgid so user changes are properly applied during execution.
• A vulnerability existed in the netty bundle’s HTTP codec (CVE-2025-58056). This issue is now fixed by updating to a patched version.

This release adds the following features:

• The agent-backend bundle now supports configurable backend read timeouts with the environment variable INSTANA_BACKEND_READ_TIMEOUT_SECONDS (default: 10; values below 10 are ignored).

• Previously, connections might reconnect too early during TLS handshakes. This is now prevented by initializing the idle state handler only after the SSL connection is fully established.

This release adds the following features:

• The sensor-api bundle provides several enhancements to process output handling:
  • Introduces a builder pattern for OsProcessOutput with configurable options.
  • Adds configurable queue capacity and batch size in StreamingOsProcessOutput.
  • Provides a reliable onClose callback for StreamingOsProcessOutput.
• The agent bundle now logs more details about network binding operations, improving troubleshooting of binding issues.
• The agent-cri-client bundle now collects both imageId and imageRef when it inspects container status, and this provides complete image metadata.
• The agent-jvm-handling bundle is refactored with the following changes:
  • Migrated from AbstractMavenArtifact to the new MavenArtifact API.
  • Now uses AbstractMavenArtifact from sensor-api instead of its own copy.
  • Removed the obsolete ByteBuddyToolsJar.java file.
• The agent-process-handling bundle improves observability and container support:
  • Logs information during network interface binding.
  • Enhances container ID detection for CRI-containerd format.
  • Improves IP resolution in Rootless Podman environments.
  • Tracks process monitoring decisions more transparently with the AgentSupportTask.
  • Provides more reliable lifecycle handling for Windows services.
• The agent-util bundle provides several enhancements:
  • A new overload of getNestedValue that returns null if a path is not found.
  • A conditional optional chain builder added to ValueFinder for more advanced value resolution logic.
  • The AbstractMavenArtifact class is now deprecated in favor of the API version.
• The agent-open-telemetry-service bundle now sorts attributes automatically by using a TreeMap, which improves performance and consistency of attribute handling.
• The agent-open-telemetry-grpc and agent-open-telemetry-http bundles now make PID lookup configurable. PID lookup is enabled by default, but it can be disabled with the environment variable INSTANA_AGENT_OTEL_GRPC_PID_LOOKUP (default: true; set to false to disable) or INSTANA_AGENT_OTEL_HTTP_PID_LOOKUP (default: true; set to false to disable). As an alternative, PID lookup can also be disabled with the configuration.yaml at:
  • com.instana.plugin.opentelemetry -> grpc -> pid_lookup .
  • com.instana.plugin.opentelemetry -> http -> pid_lookup .

• Previously, CRI-API-based process execution inside ContainerD containers failed because a required socket was missing. This issue is now fixed by adding the missing socket.

This release adds the following features:

• The sensor-api bundle now supports handling null for sense results as heartbeats.
• The agent-api bundle introduces the BindAddressProvider interface to return a list of bindable interfaces.
• The agent-cri-client now checks the default containerD socket /run/containerd/containerd.sock during startup.
• The agent-internal-api extends the ProcessInterestHandler with registration support for interfaces bound to a process.
• The agent-process-handling bundle extends the foreign PID timeout to 2 seconds and adds the n flag to mitigate DNS resolution issues.
• The agent-process-handling bundle now supports containerD through the CRI client. You can enable or disable this behavior with the environment variable INSTANA_USE_CONTAINERD_CRI_CLIENT (default: true when running in a POD).
• The agent-process-handling bundle introduces multiple new BindAddressProvider implementations:
  • AgentPublicIpBindAddressProvider : exposes the agent’s public IP as a bindable interface.
  • ContainerBridgeBindAddressProvider : exposes container bridge networks as bindable interfaces.
  • ProcessInterestBindAddressProvider : exposes process-specific network interests as bindable interfaces.
• The agent-open-telemetry-grpc bundle now allows disabling PID lookup with the environment variable INSTANA_AGENT_OTEL_GRPC_PID_LOOKUP (default is true, set to false to disable).
• The agent-open-telemetry-http bundle now allows disabling PID lookup with the environment variable INSTANA_AGENT_OTEL_HTTP_PID_LOOKUP (default is true, set to false to disable).

• Previously, annotation based logging used the generic agent-util logger. This issue is now fixed by updating all affected bundles (agent, agent-cgroup, agent-grpc-channel-cache, agent-jvm-handling, agent-process-handling, agent-runtime, agent-support-karaf, agent-task-orchestrator, agent-util, agent-open-telemetry-grpc, agent-open-telemetry-http) to use bundle aware loggers instead.
• Previously, foreign PID lookup would fail due to too short timeouts or DNS resolution. This issue is now fixed by extending the timeout and adding the n flag to mitigate DNS resolution.
• Previously, podman containers would fail to resolve host interfaces. This issue is now fixed by using host.containers.internal as a reliable host interface target.

This release adds the following features:

• The cgroup parser now exposes RSS data with total_rss for cgroupv1 and anon for cgroupv2.
• The agent-jvm-handling bundle now supports Java 25 tracing with upgraded javaagent-loader (1.3.71), instrumentation-boot (1.1.32), and instrumentation-shared (1.0.1467).
• A service is added to traverse parent processes. This service is used on various container implementations to identify the type or containerId.
• The OpenTelemetry HTTP component now allows to configure a maximum number of concurrent active channels with the environment variable INSTANA_AGENT_OTEL_HTTP_MAX_CHANNELS (default value is 1000).

• Previously, the parser for the Metadata API did not check for error strings. This issue is now fixed by checking for error responses and limiting allowed strings for the UniqueId of the agent.
• Previously, the parent process lookup might lead to an OutOfMemory error when the processes did not end with 0. This issue is now fixed by tracking already visited process IDs and stopping when a loop is detected.
• Previously, the OpenTelemetry HTTP component caused an OutOfMemory error when too many incoming requests reached the agent. This issue is now fixed by limiting the number of concurrently active channels. In addition, the expensive PID lookup is moved into the parent handler.

• Previously, the sshd process was only added to the internal list when it ended with sshd, which is not correct for newer RHEL environments. This issue is now fixed by checking how often sshd occurs in the executable and allowing when more than 1 occurrence is counted.

This release adds the following features:

• A service is added to control maximum number of concurrent reads. You can control the concurrency by using the INSTANA_INGRESS_CONCURRENCY environment variable (default value is 4).
• You can fine-tune the Netty ingress with the following environment variables:
  • INSTANA_NETTY_INBOUND_FACTOR (default value is 0.6): A value between 0 and 1, controls the percentage of direct memory that can be used for inbound.
  • INSTANA_NETTY_OUTBOUND_FACTOR (default value is 0.9): A value between 0 and 1, controls the percentage of direct memory that can be used for outbound, the outbound factor is always more than the inbound factor.
  • INSTANA_NETTY_INBOUND_SOFT_FACTOR (default value is 0.8): A value between 0 and 1, controls the percentage of direct memory that is used for a soft-backoff.
  • INSTANA_NETTY_SOFT_TIMEOUT_SECONDS (default value is 2): A value between 0 and unlimited, controls the seconds used for the soft-backoff after which memory is always granted.
  • INSTANA_NETTY_DIRECT_MEMORY_TIMEOUT_SECONDS (default value is 30): A value between 0 and unlimited, controls the seconds used for the hard-backoff after which an exception is thrown and the current message is dropped.
• The service to read configuration values can read values of type double.

• Previously, a leak in the pax-url component caused OutOfMemory errors. Now, if this issue is detected, Instana displays a monitoring issue pax_url_mvn_leak. You can fix this issue by reinstalling the agent.
• Previously, incoming ingress might starve the egress to the backend. This issue is now fixed by limiting the concurrent streams that are read concurrently to 4. In addition, the ingress also handles back-pressure by slowing down incoming data when the egress is not able to offload fast enough.
• Previously, the sshd process was only allowed when it is at the default location /usr/sbin/sshd. This issue is now fixed by relaxing the lookup to check if the process ends with sshd.

This release adds the following features:

• The service to handle exponential backoff now keeps track of the end reason when it is marked as done.

• Previously, a race condition occurred when reading the /status file, which resulted in the failure to read the in-container PID. This issue is now fixed and the agent retries reading the /status file when the content is empty. The success and error cases are added to the Collect Support Information.

This release adds the following features:

• A service is added that allows for exponential backoff.

• Previously, when a Java process stopped without a clean shutdown, files that are related to Java sensors accumulated on the user's temporary directories. This issue is now fixed by checking all known temporary directories on Windows for Java sensor files and deleting them if the Java process is stopped.
• Previously, a NullPointerException error occurred when comparing the PID with the reported PID in announce messages that are sent by tracers. This issue is now fixed. The agent always checks for null values before comparing. In addition, a backoff strategy is added to lookup the in container PID by using /status or /sched files. The number of tries and the current state of the lookup for in-container PIDs are added to the Collect Support Information.

This release adds the following features:

• Processes now report their PID as seen from the process. This information can differ between processes that run on on host and in container because the PID namespace is different.

• Previously, if inotify handles were exhausted, configuration files were not watched. This issue is now fixed by reading the metadata of the configuration files at regular intervals and updating the configuration if the metadata has changed.
• Previously, multiple processes were checked dueing the announcement of tracers. This issue is now fixed by limiting the process list to include only the processes that have the same reported PID.
• Previously, the javaagent-loader logged changes to the /tmp/.instana always on WARN level. This issue is now fixed by only logging on WARN if the rights of the folder do not match the expected level.
• Previously, the javaagent-loader failed to attach to J9 JVMs when the /proc/<pid>/tmp/.com_ibm_tools_attach was not reachable. This issue is now fixed by checking if the folder /tmp/.com_ibm_tools_attach is reachable.
• Previously, when you copied files into a folder within a rootless Podman container, a file with a random name was created. This issue is now fixed by copying the file into an intermediate folder on /tmp on the host first and keeping the original name.

This release adds the following features:

• Selective monitoring now supports Kubernetes namespaces when NAMESPACES_DETAILS_PATH is present. The Operator sets this environment variable that contains the location to a yaml file.
  • When the environment variable is set, the agent caches content for each namespace by using the instana-workload-monitoring label.
  • The following values are evaluated: <absent>, true, false, apm, and infra.
  • If INSTANA_SELECTIVE_MONITORING is set to DEFAULT or has an empty value, the agent operates in normal mode. All namespaces with instana-workload-monitoring=false are ignored.
  • If INSTANA_SELECTIVE_MONITORING is set to OPT_IN, the agent operates in the selective monitoring mode. The agent monitors only the processes that are in a namespace, which have instana-workload-monitoring=true, instana-workload-monitoring=apm, or instana-workload-monitoring=infra. All processes with instana-workload-monitoring=false are ignored.
• An internal service is added to monitor configuration file changes and notify if there is a change. This service uses the scheduler to reduce the overall thread count.

• Previously, only successful container metadata was cached. This issue is now fixed, and also a failure to look up container metadata is cached.
• Previously, cgroups, such as /init.scope and /mde, were identified as potential containers. This issue is now fixed after adding a table of known non-container cgroup names.
• Previously, ContainerD lookup failed when multiple sockets were present on the host. This issue is now fixed by using the socket that is assigned to the ContainerD process.

This release adds the following features:

• Caching Kubernetes labels for the following container types to support selective monitoring: ContainerD, CRI-O, Docker, and Podman
• Counting tasks for OTel gRPC by using the ExecutionPipeline

• Previously, the system might fail reading the configuration.yaml file if the file contained non-UTF-8 characters. This issue is now fixed by checking for non-UTF-8 characters.
• Previously, the system might display incorrect data while reading cgroup data for container metrics. This issue is now fixed by checking if any process is still active for this cgroup.
• Previously, OTel HTTP connections were cached forever. This issue is now fixed by removing closed connections from the internal cache.
• Previously, OTel HTTP connections might overload the thread pool. This issue is now fixed by only evaluating open connections.

This release adds the following features:

• You can now enable OpenTelemetry HTTP and gRPC components by using the INSTANA_AGENT_OTEL_HTTP and INSTANA_AGENT_OTEL_GRPC environment variables (default value is null).
• You can now configure default deflate level (0–9) by using INSTANA_BACKEND_DEFLATE_LEVEL variable (default value is null).
• Integer and Boolean settings (with null fallback) are now fetched from the agent environment through the new runtime and runtime-api methods.
• The agent now uses ExecutionPipeline for both JVM connections and HTTP endpoint handling in the socket-server.
• The new built-in helper ThreadWatcher monitors stuck threads. (default is 60 s. You can configure this by using the environment variable INSTANA_THREAD_WATCHER_TIMEOUT_SECONDS ).
• Tasks are now wrapped with pre/post hooks, thread-watching, pool statistics, and task monitoring by using the orchestrator-API interfaces.
• The agent now looks up nested configuration values and returns checked exceptions in callables and runnables by using the new helpers in the util bundle.
• Span-links support is added for both OpenTelemetry base and service bundles.
• OTel gRPC and HTTP components now support stuck-thread detection and task counting by using ExecutionPipeline. You can override their default ports either by using INSTANA_AGENT_OTEL_GRPC_PORT and INSTANA_AGENT_OTEL_HTTP_PORT environment variables (default value is null) or by modifying the configuration.yaml file. For more information, see OpenTelemetry configuration

• Previously, log-level synchronization with JVMs began immediately, which resulted in race conditions in internal test validation. This issue is now fixed by delaying the log-level sync start by one minute.

• Previously, when you reloaded the in-memory ClassLoader, the stored classes were removed, which resulted in missing definitions after reloading a sensor. This issue is now fixed by preserving the class-holding data structure in a global in-memory representation.
• Previously, when a Spring-Boot Java agent was started, the javaagent-loader thread was blocked if ManagementFactory was not initialized. This issue is now fixed by preloading a default ManagementFactory before starting any sensor agents in the javaagent-loader.
• Previously, when a backend connection was reset, the direct-memory guards were left in an inconsistent, non-thread-safe state, which caused the threads forwarding metrics or traces to hang. This issue is now fixed by introducing the globally shared ByteBuffer allocator with built-in timeouts.

• You can now configure a minimum heap threshold for the in‑memory classloader by using the INSTANA_AGENT_LOADER_MIN_IN_MEMORY_HEAP_MEGABYTES environment variable (default value is 100 MB), which is enabled by the the upgraded javaagent‑loader 1.3.68. Make sure that you set the environment variable in the target JVM.

• Previously, when you used the in‑memory ClassLoader on JVMs with limited heap, OutOfMemory errors occurred. This issue is now fixed by disabling the in‑memory ClassLoader when maximum heap is below the configured threshold (100 MB by default) and falling back to file‑based attachment.
• Previously, the agent issued a Unix-specific command kill -3, which failed on Windows systems. This issue is now fixed by suppressing the kill -3 call on Windows systems.

This release adds the following features:

• The agent-jvm-handling bundle now supports Java 24 tracing with upgraded javaagent-loader (1.3.67), instrumentation-boot (1.1.31), and instrumentation-shared (1.0.146).
• A new runtime service is added to inspect mount options on Unix-based systems.
• A dedicated socket-server API service is now available to start and stop Netty-based socket servers.
• The task-orchestrator’s Scheduler is now provided as a prototype, giving each consumer its own instance while using a shared global thread pool.
• The sensor API now includes a Streaming Process Output helper and JVM job-name lookup on IBM i.

This release fixes the following issues:

• Previously, sensitive fields could be exposed in the output of the process-listing (ps) task and included in support bundles. This issue is now fixed by hiding all sensitive data during support information collection.
• Previously, Support Information ZIP archives included a leading slash, which prevented file extraction on Windows. This issue is now fixed by removing the leading slash.
• Previously, determining which JVMs to ignore could enter an infinite loop when following parent-process chains. This issue is now fixed by breaking the circular lookup loop.
• Previously, applications running on J9 JVMs with custom executables or attach directories were not detected correctly. This issue is now fixed by enhancing the J9 detection logic to recognize custom paths.
• Previously, Byte Buddy attachments would fail on Unix hosts with /tmp mounted noexec, causing tracing setup to break. This issue is now fixed by detecting noexec on /tmp and falling back to JDK-based attachment.
• Previously, if container metadata was null, the agent attempted to parse invalid data and returned errors. This issue is now fixed by checking for null metadata and returning an empty map instead of parsing invalid data.
• Previously, all consumers of the Scheduler shared a single instance, which resulted in new tasks being blocked when one consumer stopped. This issue is now fixed by exposing the Scheduler as a prototype so that each consumer gets its own instance.
• Previously, shutting down the Otel gRPC socket server could hit a race condition that blocked termination indefinitely. This issue is now fixed by using the new socket-server API service to coordinate the shutdown, which ensures a stable, deadlock-free termination.
• Previously, shutting down the Otel HTTP socket server could hit a race condition that blocked termination indefinitely. This issue is now fixed by using the new socket-server API service to coordinate the shutdown, which ensures a stable, deadlock-free termination.
• Previously on IBM i, only the numeric process ID was shown instead of the full job name. This issue is now fixed by fetching and displaying the complete job name.

This release fixes the following issues:

• Previously, exceptions that occurred during the activation of the CriClient and channel caches were not captured, which led to unhandled errors during startup. This issue is now fixed by capturing and logging the exceptions during the activation of these components.

This release adds the following features:

• A dedicated support bundle for Karaf-related support collection data is added. The new agent-support-karaf bundle also adds support tasks to collect bundle tree dependencies and header information.
• The agent-jvm-handling bundle now supports a configurable timeout for JVM connections by using the INSTANA_SOCKET_CONNECTION_TIMEOUT environment variable (default value is 45 seconds).

This release fixes the following issues:

• Previously, CRI-O Sensors stopped functioning during updates because process collection was delayed. This issue is now fixed by making sure that that process collection is fully completed before proceeding with updates.
• Previously, an exception occurred during an upgrade if CRI-O containers were stopped immediately after startup. This issue is now fixed by performing a version check during client initialization to preload necessary classes.
• Previously, the use of selective monitoring sometimes resulted in incomplete data collection for containers and processes. This issue is now fixed by querying all containers and processes without relying solely on selective monitoring.
• Previously, negative values processed during startup of cgroup data led to spikes in memory and CPU usage. This issue is now fixed by normalizing any negative values to zero.
• Previously, cgroup detection was attempted on non-Linux systems and systems without a proper /proc directory, which caused errors. This issue is now fixed by restricting cgroup detection to Linux hosts with a valid /proc folder.
• Previously, memory usage for CGroupV2 was calculated inaccurately. This issue is now fixed by correctly calculating memory usage as the sum of file and anon memory.
• Previously, Windows JVM attachments used a single global TEMP directory regardless of user privileges, which caused conflicts and failing attachments. This issue is now fixed by making sure that the attach process uses the TEMP directory of the user.
• Previously, there was no mechanism to control the timeout for JVM connections, which sometimes led to connection issues. This issue is now fixed by adding an environment variable that you can use to configure the JVM connection timeout.
• Previously, failures to bind ports for ingress communications did not trigger any retry attempts, leading to connectivity problems. This issue is now fixed by implementing a retry mechanism for port binding failures.
• Previously, null values in environment variable maps during container-specific queries could lead to errors. This issue is now fixed by adding proper handling for null cases.

This release adds the following features:

• The agent framework now supports method-based trace logging. Although this feature is available through framework, currently one must annotate individual methods to enable trace logging. The full adoption is planned to be released in the future releases.
• The Log4j2 pattern is now automatically updated on startup to include the method information by using the pattern, | %replace{%X{method}}{^(.+)$}{$1 | }.
• Enhanced debug logging for processes is now available during support information collection. Every process type now provides detailed internal debug data to facilitate troubleshooting and analysis.

This release fixes the following issues:

• Previously, PID 1 was misidentified as a container process. This issue is now fixed by correctly identifying PID 1 as a host process.
• Previously, the tracer announce procedure failed on containerized workloads. This issue is now fixed by relaxing the container environment lookup for ContainerD processes and correctly handling PID 1 as a host process.
• Previously, a NullPointerException error occurred in ContainerD processes when the namespace was null. This issue is now fixed by adding a null check.
• Previously, the cgroup data lookup did not correctly utilize available processes within a container. This issue is now fixed by using any process within the container for cgroup data lookup.

This release adds the following features:

• Additional gRPC components are now exported to extend the agent’s communication capabilities.
• The agent configuration steps are run sequentially in a predefined order to improve the configuration process.
• The agent can now replace configurations from both environment variables and files within the agent context.
• Automatically merges sensor data from sources that share the same name and entityId.
• The agent now provides selective monitoring for container environments.
• Activates the internal CriClient by default, ensuring seamless integration with modern container runtimes.

This release fixes the following issues:

• Previously, the JVM attachment process would fail on J9 JVMs when any of the environment variables (IBM_JAVA_OPTIONS, JAVA_TOOL_OPTIONS, or _JAVA_OPTIONS) defined the com.ibm.tools.attach.directory explicitly. This issue is now fixed by evaluating these environment variables and updating the attach-tools (v1.17.2) to correctly handle an explicit directory definition.
• Previously, when the host had a large number of processes with identical arguments and environment variables, the agent would run out of memory. This issue is now fixed by interning strings, arguments, and environment variables, which reduces the memory footprint.
• Previously, a file permission issue occurred when copying files into a rootless Podman container under strict umask settings. This issue is now fixed by applying file permissions to the intermediate file before it is copied.
• Previously, in Red Hat OpenShift Container Platform 4.18 environments where crun is the default container runtime, the agent did not use the internal CriClient by default. This issue is now fixed by handling INSTANA_USE_CRI_CLIENT as true if the variable is absent, which ensures compatibility with the new container runtime default.
• Previously, a jail root lookup issue on Solaris caused all processes to be incorrectly identified as jail root. This issue is now fixed by restricting jail root process handling to Linux systems only.

This agent release adds the following features:

• A new method is added to query all known process IDs (PID) for process‑based sensors. This new method improves the tracking and monitoring of processes.
• The process handling bundle now uses an event‑based update mechanism instead of the previous timer‑based approach to update iptables in Garden-based containers.

This release fixes the following issues:

• Previously, the cgroup bundle conducted unnecessary cgroup version checks on non‑Linux systems. This issue is now fixed by limiting the version check to Linux environments.
• Previously, the memory parsing process in the cgroup bundle added wrong limits. This issue is now fixed by adding two new helpers, parseMemoryDataCgroupV1 and parseMemoryDataCgroupV2 to read the correct memory files.
• Previously, the JVM handling bundle reported a jvm_tmp_not_available monitoring issue when the /tmp directory was unavailable for jail root-based JVMs. This issue is now fixed by updating the monitoring issue to agent_tmp_directory_missing and adding support for the continuous monitoring of jail root JVMs.
• Previously, the exceptions that were identified in the symbolic link checks on Linux were logged at a high level of detail, which cluttered the logs. This issue is now fixed by reducing the log level to DEBUG.
• Previously, the process handling bundle encountered issues with normalizing broken paths on Windows. This issue is now fixed by enhancing the logic for Windows path normalization.

This agent release adds the following features:

• Support for jail root is added for both JVMs and processes.
• Handling of the capabilities endpoint in the agent backend is enhanced.
• OpenTelemetry integration is improved by linking the host and OpenTelemetry entities by using MachineId.

This release fixes the following issues:

• Previously, instrumentation logs experienced a memory leak. This issue is now fixed by updating instrumentation-shared files to 1.0.144.
• Previously, the gRPC server started as part of the agent grpc bundle even when it was not required. This issue is now fixed by moving the the gRPC server from the agent grpc bundle to the agent open telemetry grpc bundle.

This release fixes the following issues:

• Previously, the location of ctr on Tanzu Clusters was not identified correctly, which caused issues with container operations. This issue is now fixed by adding the logic to accurately detect the ctr location in such environments.

This agent release adds the following features:

• The UnifiedProcess abstraction is added to offer extended functionalities to manage normal processes and CRI processes.
• CriClient and its associated configurations are added for seamless interaction with CRI API gRPC endpoints.
• The agent cgroup bundle is added to enhance container metric management.
• The agent util bundle now provides improved resource and thread management.

This release fixes the following issues:

• Previously, entities in the backend were marked as offline due to missing data. This issue is now fixed by keeping entities alive with empty data.
• Previously, the agent's handling of empty files caused unexpected behavior. This issue is now fixed by returning a zero-length byte array when empty files are encountered.
• Previously, the asynchronous process for adding public IP gateway interests caused delays. This issue is now fixed by changing the operation to be synchronous.
• Previously, in rootless Podman environments, temporary files that were created with the correct set of permissions were overwritten because of the umask settings. This issue is now fixed by explicitly setting the file permissions after the temporary file has been created.

This agent release adds the following features:

• The Instana agent has stopped the support for Java 8. You must use Java 11. When you start newly installed agents with Java 8, the agent now shows a monitoring event. This monitoring event is relevant only if you install or use your own JVM. Containerized or package-based agents already use Java 11.

This release fixes the following issues:

• Previously, in rootless Podman environments, copying files to the target container might fail as the agent did not define a permission set for the temporary file. This issue is now fixed, and the temporary files are created with rwxr-xr-x.
• Previously, the agent failed to set any socket information for a ContainerD instance when multiple sockets are present. This issue is now fixed, and the agent sets the correct socket information for a ContainerD instance.
• Previously, the agent had a race condition on JVM connections and kept a connection open even though it was internally already marked as closed. The wrong connection was removed only when the agent was restarted. This issue is now fixed, and the agent correctly closes the broken connection.

This release fixes the following issues:

• Previously, in rootless Podman environments, the df command failed a check that is used to verify the available size within a container before a JVM attachment. This issue is fixed, and the df command is now run by the user as required.
• Previously, during the startup, the agent failed to execute some commands that are necessary to identify if the agent is running on a host that supports various container environments. The failure occurred because the default timeout to run the command was one second, which was not sufficient. This issue is fixed, and you can now control the timeout by using a new environment variable INSTANA_COMMAND_RUNNER_TIMEOUT_SECONDS=5, which has a default duration of 5 seconds.
• Previously, the agent did not include the --user argument while executing a nsenter command when the environment variable NSENTER_MODE_ENABLED=true was set. This issue is fixed, and the agent now uses the --user argument when required.

This release fixes the following issues:

• Previously, the uniqueId from the agent on Linux was not deterministic when the main interface identified by Sigar was a vlan interface. This was because these interfaces are not ordered by the host. This issue is fixed, and the vlan interfaces are now sorted first if names that start with vlan are identified.
• Previously, the uniqueId from the agent on Windows used the KM-TEST Loopback Adapter interface. Because this is a loop-back interface, the MAC address was not unique. This issue is fixed, and the agent now ignores the KM-TEST Loopback Adapter interface when other interfaces are available.
• Previously, the agent did not log when the agent key for a backend was wrong. This issue is fixed, and the agent now logs a warning every 10 minutes when it receives a 403 error report from the backend, which indicates that the agent key is wrong.

This agent release adds the following features:

• The agent now has a Selective Monitoring mode. This mode adds the ability to monitor only the processes that hold specific environment variables.

  The agent checks if the process is started with the environment variable INSTANA_SELECTIVE_MONITORING.

  • If INSTANA_SELECTIVE_MONITORING is set to DEFAULT or has an empty value, the agent operates in normal mode. All processes with INSTANA_MONITORING=false and INSTANA_IGNORE=<any value> are ignored.

  • If INSTANA_SELECTIVE_MONITORING is set to OPT_IN, the agent operates in Selective Monitoring mode. The agent monitors only the processes that have INSTANA_MONITORING=true, INSTANA_MONITORING=apm, or INSTANA_MONITORING=infra. All processes with INSTANA_MONITORING=false and INSTANA_IGNORE=<any value> are ignored.

  Currently, there is no differentiation between apm or infra modes.

This release fixes the following issues:

• Previously, the agent always executed the id command in the context of the agent, which could fail if the agent was running in a container. This issue is fixed, and the id command is now executed on the host.
• Previously, the agent failed to start because of Snakeyaml imports that required a minimum version of 2. This version was not present in the agents with agent bootstrap version earlier than 1.2.9. This issue is fixed, and the agent util now imports Snakeyaml versions 1 to 3.

This release fixes the following issue:

• Previously, the agent only checked if the /etc/passwd file contained a user ID to identify a user. This issue is fixed, and the agent now correctly identifies a user by using the id command instead.
• Previously, the agent had a fallback setting to use the sudo command on AIX for the JVM attachment. However, the fallback was not always possible as the sudo command is not available on AIX by default. This issue is fixed, and the agent now only uses the su command on AIX for the JVM attachment.

This agent release adds the following features:

• The agent now has a dedicated bundle to handle all activities that are related to processes. These activities also include the abstraction around containerized processes.
• The agent now collects a validation of the confiugration.yaml. You can retrieve this data by clicking on the Collect Support Information button on the agent dashboard.

This release fixes the following issues:

• Previously, the agent did not check if the /etc/passwd file had more characters to evaluate when a user ID was not found in the file. This issue is now fixed, and the agent correctly identifies when the checking of the file is finished.
• Previously, the agent did not evaluate all the available namespaces of the host system when executing the nsetner -a command. This issue is fixed, and the agent now checks all available namespaces and adds them explicitly to the nsenter command.

This agent release adds the following features:

• The agent now collects a thread dump of all agent threads. You can retrieve this data by clicking on the Collect Support Information button on the agent dashboard.

This release fixes the following issue:

• Previously, the agent did not use the correct PATH when it was run in a container on RKE2. This issue is now fixed, and the agent correctly identifies when it is run on RKE2. The agent also adds /var/lib/rancher/rke2/bin to the PATH if it is missing.
• Previously, the agent used a thread from the scheduler during rebooting or resetting. When the agent installed other features at the same time, the agent could get stuck on the reboot or reset. This issue is now fixed, and the agent uses a thread from the main group and finishes the reboot or reset.
• Previously, the agent used the su command to change the user on IBM i when the agent was run as the root user QSECOFR. This attachment failed because IBM i does not support su. This issue is now fixed, and the agent uses the custom ibmi-sudo executable to change the user.
• Previously, the agent used the sudo command to change the user on z/OS as a fallback. This attachment failed because z/OS does not support sudo. This issue is now fixed, and the agent only uses the su command to change the user.
• Previously, the agent did not provide details about the threads that are stuck when the scheduler and executor stopped the threads for the agent to stop, restart, or update. This issue is now fixed, and the agent writes a thread dump into the logs when the threads are stopped.
• Previously, the agent attached to JVMs that are part of a buildah step. This issue is now fixed, and the agent ignores JVMs that have a parent process with the name buildah-chroot-exec.

This release fixes the following issue:

• Previously, the agent returned a NumberFormatException error in the PodmanUtil service and failed to start most of the internal services on AIX systems. This error occurred because the UIDs on AIX exceeded the maximum integer range that was specified in the Agent 1.2.6 release. This issue is now fixed, and the agent parses long UIDs and starts up correctly.

This agent release adds the following features:

• The agent now supports rootless Podman to identify if a process is started by the root or a user. To use this feature, make sure that Podman sockets are available for both root and user.
• The agent now provides an abstraction layer to access files for rootless Podman. The abstraction is required because the existing Java Nio-based file access has the same user as the agent and therefore cannot interact with the user file system.
• You can now configure the agent to send less data to the backend from the sensors that collect data at lower granular intervals. For example, process sensors and container sensors. Decreasing the amount of data that is sent to the backend reduces the counted traffic and the computation on the backend.
• The agent now generates lesser number of metrics on the backend to debug traces and logs. This reduces the load on the backend.

This release fixes the following issues:

• Previously, the agent marked a process as not running after an update of the agent feature. This was because the configuration service that is used to determine whether a process must be ignored is not ready at startup time. This is now fixed and the agent awaits the configuration service before deciding whether a process is ignored.

This release fixes the following issues:

• Previously, the agent failed to run a docker exec command with a user switch. This was because the agent added the -u flag after containerId, which is invalid. This issue was noticed in the Agent 1.2.3 release. This issue is now fixed.
• Previously, the agent created a jvm_instrumentation_shared_absent monitoring issue when a JVM was absent or when the connection to the JVM was already closed. This issue is now fixed and the agent creates the jvm_instrumentation_shared_absent monitoring issue when the JVM is still active.

This release fixes the following issues:

• Previously, the agent became unresponsive if multiple restart requests were raised at the same time. This issue is now fixed and the agent ignores incoming restart requests in the first minute.
• Previously, the agent did not check whether the instrumentation-shared files exist. If the files were absent, the java-tracer did not start and a NoSuchFileException was added to the logs. This issue is now fixed and the agent checks if the instrumentation-shared file is present. The agent also creates a monitoring issue with the name jvm_instrumentation_shared_absent if the instrumentation-shared is absent.
• Previously, the agent generated multiple logs when an update occurred while the agent was forwarding logs to the backend. This issue is now fixed and logs are not created when the agent is updated.

This agent release adds the following features:

• The agent now collects the configurations that are shared with a sensor at startup time or as part of a configuration update. You can retrieve this data by clicking on the Collect Support Information button on the agent dashboard.
• The agent now supports JDK-based attachments along with Byte Buddy-based attachments. You can specify to use JDK-based attachments by setting the environment variable INSTANA_JVM_BYTEBUDDY_ATTACHMENT_ENABLED as false.

This release fixes the following issue:

• Previously, the agent created a new string when comparing the inventory as part of the discovery check. This issue is now fixed and new strings are not created anymore.

This release fixes the following issue:

• Previously, the vault configuration integration failed to start and became unusable when the agent bundle was updated after the vault was started. This issue is now fixed, and the vault configuration integration works after an update.

This release splits the agent bundle into multiple bundles. The splitting of bundles reduces the duration of agent downtimes during an update and ensures that the components that are not updated are run without interruption. This agent release contains the following bundles:

• agent 1.1.729: The main agent bundle.
• agent backend API 1.0.0: Services required for other agent bundles for backend connection.
• agent backend 1.0.0: The implementation of the backend connection.
• agent backend util 1.0.0: Utility bundle for backend-related components.
• agent internal API 1.0.8: Services required for other bundles that are provided by the agent bundle.
• agent Jaeger tracer 1.0.0: The Jaeger Tracer component.
• agent memory API 1.0.0: Services required for other bundles for memory access.
• agent memory 1.0.0: Implementation of the memory access.
• agent runtime API 1.0.0: Services required for other bundles of the agent that allow access to the runtime-related data.
• agent runtime 1.0.0: Implementation of the runtime-related data.
• agent self profiling 1.0.0: Self-profiling capability of the agent.
• agent socket server API 1.0.0: Services required for other bundles of the agent for the main agent socket server 42699. Mainly holds counter related information.
• agent socket server 1.0.0: The implementation of the main agent socket server that runs on port 42699.
• agent task orchestrator API 1.0.0: Services required for other bundles for task orchestration.
• agent task orchestrator 1.0.0: The implementation for the task orchestration. Holds the Scheduler and Executor services.
• agent util 1.0.0: Helper bundle with static helpers that are not part of any API.
• agent Zipkin tracer 1.0.0: The Zipkin Tracer component.

This agent release adds the following features:

• Java Instrumentation now supports Java 23.
• The agent now supports a special environment variable INSTANA_BIND_ALL_INTERFACES. If this environment variable is set to true, the agent binds to all interfaces and announces the interfaces when the agent monitors Java workloads. This feature is required when the container uses Macvlan.
• The agent now tracks aborted requests on the 42699 endpoint for internal monitoring.

This release fixes the following issues:

• Previously, an internal service of the agent would be absent if a temporary directory on the host was corrupted. Now, the agent catches this exception and the internal service remains intact.
• Previously, the agent would time out on the java -version check after 5 seconds. This timeout is now increased to 10 seconds.

This agent release adds the following feature:

• The agent now runs on AS400/IBM i OS.

This release fixes the following issues:

• Previously, the agent showed the agent-bundle version on the / path. This issue is now fixed, and the agent now displays the agent version.
• Previously, the agent ignored connections from JVMs if they were marked as unstable. These markings did not timeout. The timeout for the markings is now fixed, and the potential unstable connections are retried again after 5 minutes of the timeout.

This agent release adds the following feature:

• An internal API is added to the host agent that enables the sensors to request for the process environment configuration that is already in discovery.

This release fixes the following issue:

• Previously, J9-based JVMs were excluded from the new attachment that was added with Agent 1.1.721 because these JVMs required the stat command, which was not available on AIX. This issue is now fixed and the JVMs are included in the new attachment because the stat command is replaced with the istat command.

This release fixes the following issues:

• Previously, the Host Agent added a dedicated Thread Pool for discovery requests. An edge case was identified when the Agent was receiving too many discovery requests on startup. In that case, the Host Agent might not handle any other Request and failed to respond to health checks. This overload might lead to restarts on Kubernetes. This issue is now fixed and the Host Agent is no longer restarted.
• On Solaris, the Host Agent checks if the <java-home>/jvm/jre/lib/sparcv9/jspawnhelper is executable. Now, the Host Agent notifies the user with a Monitoring event.
• Previously, the Host Agent showed the command that is used on the java -version process in an error case. As this command can display secrets, the used command is no longer shown.

• Previously, Opentelemetry monitoring was deactivated by default. The new default is activated. For more information, see Configuring OpenTelemetry data ingestion.

This release fixes the following issue:

• Previously, the Host Agent verified incoming data on the connection and closed it if no traffic was observed within 600 seconds. An edge case was identified when the Agent loses connection to the Instana backend every 30 seconds. Closing the connection led to extra load within the JVM. This issue is now fixed and the Host Agent now checks the incoming data at 200, 400, and 600 seconds. If no data reaches the agent for 600 seconds, the connection is marked as unstable and closed.

This release fixes the following issues:

• Previously, the Host Agent ignored a process after reading the content in /proc/<pid>/<something> had a timeout. The default timeout was 5 seconds, which got triggered mistakenly in Host Agents running on overloaded hosts. In such cases, the Agent mistakenly ignored the process and did not retry it. This issue is now fixed, and the Host Agent handles potentially broken processes with a staged approach. The Host Agent now supports multiple stages for a process. Successful reading of process data resets the counter. The list of the ignored processes is reset every 10 minutes.
• Previously, in some cases within the Host Agent, the flow expected a connection to be removed to run a full reattachment. This reattachment is only possible if the connection is removed and is not added back within a timeout of 20 seconds. This issue is now fixed, and the connection between the Host Agent and JVM is now removed, if the connection is marked as closed.
• Previously, a regression with J9 based JVMs occured that required the stat command, which was not available on AIX. This issue is now fixed and now the JVMs that run on AIX are excluded from the new attachment that is added with Agent 1.1.721. The new attachment is enabled again when stat command is replaced with istat command.

This release fixes the following issues:

• An environment variable CONTAINERD_SUPPRESS_DEPRECATION_WARNINGS with the value true is added to the ctr exec operations. This environment variable is to exclude deprecation warnings that might interfere with parsers in the Agent.
• The evaluation of a process and the PID that reaches the 42699 port of the Agent is delayed. Computation is completed when the field is requested in code. When you make computation optional, the Agent executes other important tasks.

This release fixes the following issues:

• Process with Arguments on Windows were ignored if they contained characters outside of the UTF-8 area.
• The Javaagent Loader inside a JVM leaks on Exit Hooks for file references of the Agent when the attachment works, but the communication to the Agent failed afterward
• When the agent attaches to a JVM, the Agent places files in <TMPDIR>/.instana with read and execute permissions and excludes the write permission for everyone.
• The Agent now parses the output of docker info when Docker is available to save the location of the Docker Root Dir. This information is used to correctly identify Docker containerIds.

This release fixes the following issues:

• Better logging when java -version fails. The log now contains the command and the output that failed to be parsed.
• JVM attachments now always use the attach-tools-<version>.jar, which is placed in <TMPDIR>/.instana to allow attachments against JREs and 32-bit JVMs on Windows.
• The Agent monitors the Metaspace and issues a restart if the available space is under 10 MB for over 30 minutes. This restart reduces the risk of an OutOfMemoryError:Metaspace.

This release fixes the following issue:

• JVM monitoring: Ignore SAP NetWeaver classes to mitigate class loading issues.

This release fixes the following issues:

• JVM monitoring: Support for Virtual Threads.
• Reduction in Memory by deduplicating Strings on internal tasks.
• Vault is available to other Sensors by an internal Service.