Service, status, and logs

The Linux agent runs as a systemd service named cloudstats-agent.

Service#

Unit file: /etc/systemd/system/cloudstats-agent.service

Verified settings from that unit:

• Starts after network-online.target
• Restart=always with RestartSec=5
• RestartPreventExitStatus=87 (used when a deleted server is committed locally; see Uninstall the agent)
• TimeoutStopSec=20
• Runs as root
• ProtectSystem=strict and PrivateTmp=true
• Write access only to /etc/cloudstats and /var/lib/cloudstats

Commands:

systemctl status cloudstats-agent
journalctl -u cloudstats-agent

The website treats a server as online when the agent has made authenticated contact within the last 60 seconds. That is contact with CloudStats, not a guarantee that the latest chart point was stored.

Collection interval#

The agent collects on an interval stored in /etc/cloudstats/agent.json as next_interval_seconds. The default is 10 seconds. The agent clamps values into 5–3600 seconds.

One collection cycle writes one batch. CPU percents and byte-per-second rates need a previous sample, so they are omitted from the first cycle for that series. See Current metrics catalog.

If the host cannot reach CloudStats, batches stay in a local queue under /var/lib/cloudstats/wal/ (at most 64 MiB or 512 files). Replay is oldest-first. Charts that read the latest snapshot can lag until that queue drains.

Files on the host#

Logs#

The agent writes JSON logs to stdout. systemd captures them:

journalctl -u cloudstats-agent

Successful start logs agent running with server_id and agent_id. Enrollment logs those ids and does not log the enrollment token or the permanent credential.

Website status versus charts#

If the agent is online but charts are stale, the host may be retrying a backlog or CloudStats may be accepting contact without yet accepting a batch. See Troubleshooting.