HPA Portal sections Data collection Data collection session troubleshooting

The Data collection session is a critical and sensitive process.

The first reference to troubleshoot a data collection session is the standard HPA troubleshooting process.

Simultaneous data collections

What is a simultaneous data collection?

Each HPA agent can run only one data collection session at a time. When a new session is started while some agents are already participating in another session (e.g. launched by another user, on another environment, or triggered by an alert), those agents reject the new request.

This situation is called a simultaneous data collection. The rejected agents do not contribute data (KPI, logs, thread dumps, etc.) to the new session — their data remains in the session that was already running.

The simultaneous_data_collection.txt file

When at least one agent is rejected, a file named simultaneous_data_collection.txt is included in the data collection session archive (zip). This file contains:

  • A warning indicating the session was impacted
  • The list of data types that are missing for the rejected agents (KPI, logs, thread dumps, memory histograms, heap dump, JFR recording — depending on the session configuration)
  • The list of rejected agents, identified by host, process, and agent name
  • For each rejected agent, the details of the conflicting session: name, environment, user, and start time

How to find the missing data

The data for the rejected agents is available in the archive of the conflicting session listed in the flag file. To retrieve the complete picture:

  1. Open the simultaneous_data_collection.txt file in the archive
  2. Note the conflicting session name, environment, and user
  3. Locate the archive of that session in the Explore archived sessions tab
  4. The missing agents’ data (KPI, logs, dumps, etc.) is in that archive

How to avoid simultaneous data collections

  • Before starting a data collection session, check the In progress tab to see if another session is already running on the target agents
  • Coordinate with other users when multiple data collection sessions may be needed on the same environment
  • When a data collection is triggered by an alert action, be aware that manual sessions on the same agents will conflict

JVM shutdown during a data collection session

What happens when a JVM stops during a data collection?

If a monitored Calypso JVM stops while a live data collection session is running, the session on that JVM is interrupted. The data collected up to that point (KPI, logs, thread dumps, JFR recordings) remains on the agent host but has not yet been transferred to the EM.

Automatic recovery

HPA automatically recovers interrupted sessions. The data collection session monitor periodically checks for sessions left behind by a JVM that is no longer running. When such a session is detected:

  1. The monitor transfers the collected data to the EM
  2. The EM rebuilds the archive with the recovered data
  3. The session appears in the Explore archived sessions tab as usual

No manual intervention is required. The recovery runs in the background and typically completes within a few seconds after the JVM stops.

What to expect in the archive

A recovered session archive contains data collected up to the moment the JVM stopped. Depending on when the interruption occurred, some data types may be partial or shorter than the originally requested duration.

Limitations

  • Automatic recovery is only available on Linux hosts
  • Recovery requires the data collection session monitor to be running on the same host (see Data collection session monitor)