Connectivity & Resilience

Edge devices run unattended, often in the field, on home or factory networks you don’t control. The IronFlock device agent is built to keep every device online and remotely reachable — so you can always reach it to diagnose and fix problems, even when the device itself is under stress.

The agent treats remote reachability as the top priority. Whatever goes wrong locally — the network drops, the disk fills, memory runs out — the agent’s job is to stay connected and manageable rather than silently fall off the grid.

Always-on network connectivity

Headless devices on DHCP networks routinely face transient outages: the router reboots, the DHCP lease hiccups, a switch power-cycles, or a cable is briefly unplugged.

The agent configures the device’s network manager to retry connectivity indefinitely. Default operating-system settings often give up after a few failed attempts and leave the interface down until someone physically reboots the device. IronFlock devices instead keep trying until the link returns — so a momentary outage never turns into a permanent disconnection.

When a device does change address (for example a new DHCP lease after a power cut), the agent re-resolves its own host details on restart, so it comes back cleanly without manual intervention.

Stays reachable under disk pressure

A full disk is one of the most common ways an edge device becomes unresponsive. The agent guards against this in three layers.

1. Prevention

The agent caps the two log sinks that most often fill a disk — container logs and the system journal — so routine operation can’t slowly consume all available space.

2. Automatic cleanup

The agent continuously watches free disk space. As space runs low, it automatically reclaims it safely — pruning unused images and build cache and trimming oversized logs. It never deletes data your apps depend on: tagged images and app volumes are always left untouched.

3. Storage Emergency state

If free space becomes critical despite cleanup, the device enters a Storage Emergency state:

• Application workloads are stopped to immediately halt further disk growth.
• New app starts, builds, and downloads are blocked so nothing can make the situation worse.
• The state is reported to the cloud and shown as a prominent marker on the device in your dashboard, so you know exactly why apps aren’t running.
• The device stays online and remotely manageable — you can still open a remote terminal and reconfigure the network to investigate and free space.

The most important property: a critically full disk degrades gracefully into a clearly-signalled, recoverable state instead of taking the whole device offline.

Automatic recovery

The agent checks frequently while in the emergency state. As soon as enough space is freed, the device automatically leaves the Storage Emergency state and restores your apps to their previous state — no manual reboot or redeploy required.

You can see a device’s live storage usage and current status in Device Settings. System-resource usage bars turn red as a resource approaches its limit, giving you early warning before a device reaches the emergency threshold.

Survives memory pressure

Low-memory conditions can hang an entire device, making it unreachable just when you need to intervene. IronFlock devices run an out-of-memory protection service that steps in under memory pressure and stops a single runaway process — keeping the agent, remote access, and the rest of the system alive instead of letting the whole device freeze.

Self-healing by design

• The agent runs as a native system service that restarts automatically if it ever exits, so the device re-establishes its connection on its own.
• It self-updates over the air, so connectivity and resilience improvements roll out across your fleet without re-flashing devices.
• Remote-access tunnels reconnect automatically after a network interruption.

Why it matters

The result is a fleet you can rely on reaching — so an unattended device in the field can almost always be reached and recovered remotely, without a site visit.