Fleet Management

The Fleet page gives you a centralized view of every agent node enrolled in the current environment. From here you can monitor node health, inspect system resources, view pipeline metrics, trigger agent updates, and stream live logs.

Node list

All enrolled agent nodes are displayed in a table with the following columns:

If no agents have enrolled yet, the page shows a prompt directing you to generate an enrollment token in the environment settings.

Node health statuses

Agent nodes report their health through periodic heartbeats. VectorFlow derives the following statuses:

• Online -- The agent is sending heartbeats within the expected interval. The node is healthy and processing pipelines.
• Unreachable -- The agent has missed heartbeats beyond the configured threshold (default: 3 missed intervals). This typically means the agent process has stopped, the host is down, or there is a network issue.

The heartbeat threshold is calculated as fleetPollIntervalMs * fleetUnhealthyThreshold. With the default settings of a 15-second poll interval and a threshold of 3, a node is marked unreachable after approximately 45 seconds of silence.

Node detail page

Click a node name to open its detail page, which provides deep visibility into that specific agent.

Node details card

A summary card shows key information at a glance:

• Status -- Current health status
• Environment -- Which environment the node belongs to
• Agent Version -- The installed VectorFlow agent version
• Vector Version -- The Vector binary version
• Last Heartbeat -- Timestamp of the most recent heartbeat
• Enrolled -- When the agent first enrolled
• Host / API Port -- Network address details
• Last Seen / Created -- Timestamps for tracking node lifecycle

System resources

Charts display real-time and historical metrics for the node's host machine:

• CPU usage -- Derived from cumulative CPU seconds
• Memory usage -- Used vs. total memory
• Disk usage -- Filesystem used vs. total bytes
• Load averages -- 1, 5, and 15-minute load averages
• Network I/O -- Bytes received and transmitted
• Disk I/O -- Bytes read and written

You can adjust the time window (up to 168 hours / 7 days) to view historical trends.

Pipeline metrics

A table shows every pipeline deployed to the node along with live throughput data:

Logs

A live log stream from the agent, with filtering options:

• Log level -- Filter by severity (DEBUG, INFO, WARN, ERROR)
• Pipeline -- Scope logs to a specific pipeline running on the node

Logs are paginated and load on demand.

Agent updates

When a newer agent version is available, an Update available badge appears in the node list. The update mechanism depends on the deployment mode:

Update errors

When a binary agent update fails (for example, due to a download error or checksum mismatch), the fleet page shows an Update failed badge on the affected node. Hover over the badge to see the error details.

Common causes:

The error badge clears automatically after the next successful update.

Pipeline deployment matrix

Below the node list, the Pipeline Deployment Matrix shows a grid of all deployed pipelines across all nodes in the environment. This lets you see at a glance which pipelines are running on which nodes and their current status.

Node labels

Labels are key-value pairs you can attach to nodes for organization and selective deployment. Common uses include tagging nodes by region, role, tier, or any custom dimension relevant to your infrastructure.

Viewing labels

In the fleet table, nodes with labels show a compact "N label(s)" button in the Labels column. Click the button to open a popover displaying all labels as key=value badges. Nodes with no labels show an empty column.

Adding and editing labels

Agent-reported labels

Agents can also report labels in their heartbeat payload. When a label is reported by the agent and also set via the UI, the UI value takes precedence. This lets you override agent-reported labels without them being overwritten on the next heartbeat.

Selective deployment with labels

When deploying a pipeline, you can optionally restrict deployment to nodes matching specific labels. In the deploy dialog, the Target Nodes selector lets you pick from all labels in the environment. Selected labels are combined with AND logic -- a node must have all selected labels to receive the pipeline.

The deploy dialog shows a live count of matching nodes (e.g., "3 of 5 nodes match") so you can verify your selection before deploying. When no labels are selected, the pipeline deploys to all nodes in the environment (backward compatible).

Node groups

Node groups let administrators segment their fleet into logical clusters based on node labels -- for example by datacenter, role, or region. Groups are managed from Settings > Fleet.

Each node group has:

Label compliance

When node groups define required labels, the fleet list displays a Non-compliant badge next to any node that is missing one or more of those labels. This is a warn-only indicator -- non-compliant nodes continue to receive heartbeats and deployments normally.

To resolve a non-compliant node, add the missing labels via the node detail page or ensure the node enrolls with matching labels so that group templates apply automatically.

Fleet health dashboard

The Health tab on the Fleet page provides an aggregated view of fleet status organized by node group. This is especially useful for large fleets where you want to see health at a glance before drilling into individual nodes.

Group summary cards

Each node group is represented as a collapsible card showing three metrics:

Drill-down

Click any group card to expand it and see a per-node detail table with:

• Name — the node name, linked to its detail page
• Status — current health status badge (Healthy, Degraded, Unreachable, Unknown)
• CPU Load — the 1-minute load average from the latest heartbeat, or -- if no metrics are available
• Last Seen — how long ago the node last sent a heartbeat
• Compliance — whether the node has all required labels for the group

Nodes are sorted by health status with the least healthy nodes shown first, then alphabetically by name.

Filtering

The toolbar above the group cards supports three filter types:

• Group — show only a specific group card
• Labels — filter by label key/value pairs (applied to the per-node detail table inside expanded cards)
• Compliance — toggle between All, Compliant (100% compliance rate), or Non-compliant (below 100%)

Ungrouped nodes

Nodes that do not match the criteria of any defined group appear under an Ungrouped card. This card behaves the same as any other group card — you can expand it to see the per-node table.

Maintenance mode

Maintenance mode lets you temporarily stop all pipelines on a node without removing it from the fleet. This is useful for host upgrades, kernel patches, disk maintenance, or any situation where you need the node idle but still connected.

Entering maintenance mode

You can toggle maintenance mode from two places:

• Fleet list -- Click the Maintenance button in the node's row.
• Node detail page -- Click the Enter Maintenance button in the header, or the Exit Maintenance button in the orange banner.

Both locations show a confirmation dialog before entering maintenance mode. The dialog warns that all running pipelines on the node will be stopped.

What happens in maintenance mode

When maintenance mode is enabled on a node:

1. The node's status badge changes to an orange Maintenance indicator with a wrench icon.
2. On the next poll, the agent receives an empty pipeline list from the server, causing all running pipelines to stop gracefully.
3. The agent continues sending heartbeats, so the node remains visible and manageable in the fleet UI.

Exiting maintenance mode

Click Exit Maintenance from the fleet list or the node detail page. No confirmation is required. On the next poll cycle, the agent receives its full pipeline configuration again and automatically restarts all pipelines.

Drift Detection

Drift detection monitors your fleet for discrepancies between the expected state (what the server says should be running) and the actual state (what agents report). VectorFlow tracks two types of drift:

Version drift

Version drift occurs when one or more nodes are running a different pipeline version than the latest deployed version. For example, if Pipeline A is at version 5 on the server but a node is still running version 4, that pipeline has version drift on that node.

Version drift is evaluated fleet-wide across all deployed pipelines in an environment. The system compares each node's reported pipeline version (from its heartbeat) against the latest deployed version in the database.

Config drift

Config drift occurs when a node's running configuration checksum does not match the expected checksum computed by the server. This can happen when:

• A deploy was initiated but the agent has not yet picked up the new configuration
• The agent's local config file was modified outside of VectorFlow
• A network issue prevented the agent from receiving the latest config

Config drift is evaluated per-node during heartbeat processing by comparing the agent-reported configChecksum against the server-side expected checksum.

Drift alerts

Drift metrics feed into the fleet alerting system. When version drift is detected, alerts can be triggered based on your configured alert rules. The alert message includes details about which pipelines are drifted and what versions each node is running.

Responding to drift

When drift is detected:

1. Version drift -- Check if the affected nodes are reachable and polling. Most version drift resolves automatically on the next agent poll cycle. If it persists, verify the agent is running and can reach the server.
2. Config drift -- Verify that the agent is receiving the latest configuration. Re-deploying the pipeline can force a fresh config push. If the agent's local config was manually modified, the next poll cycle will overwrite it with the server's version.

Node management

From the node detail page you can:

• Rename -- Click the node name in the header to edit it inline.
• Revoke Token -- Revokes the node's authentication token, preventing it from communicating with the server. The node is marked as unreachable.
• Delete Node -- Permanently removes the node record from VectorFlow. This does not stop the agent process on the remote host.