Lifecycle and observability

Start, stop, restart. Status, logs, and the live TUI dashboard.

Shortest commands

mxnode start --all          # bring everything up
mxnode status               # one-shot per-node health snapshot
mxnode logs --node 0 -f     # tail node-0 logs

Every lifecycle command takes the same selector grammar so muscle memory transfers.

Selectors

Selector	Meaning
`--all`	Every node in `state.toml` (and the proxy, if installed)
`--node N`	Just node N. Repeat for multiple: `--node 0 --node 2`
`--shard <0\|1\|2\|metachain>`	Every node assigned to that shard
`--validators-only`	All validator-role nodes
`--observers-only`	All observer- and multikey-role nodes
`--select '<expr>'`	Free-form: `role=validator AND shard=0`

Selectors are mutually exclusive — clap rejects more than one at parse time.

Start / stop / restart

mxnode start --all
mxnode stop --node 1
mxnode restart --shard 0

start maps to systemctl enable --now <unit>, stop to systemctl stop, restart to a stop+start cycle.

Restart strategy

Flag	Default	Effect
`--strategy rolling\|parallel`	`rolling`	Rolling: one at a time, wait for `is-active`. Parallel: up to `--max-parallel` concurrent.
`--max-parallel N`	`1`	Cap when `--strategy parallel`.

# Rolling (default)
mxnode restart --all

# Parallel up to 2
mxnode restart --all --strategy parallel --max-parallel 2

Status

mxnode status

mxnode 0.8.1 │ mainnet │ 4 nodes │ tags: config v1.11.5.0 binary v1.11.5
H │ idx │ unit                    │ shard      │ nonce      │ pubkey       │ port
──┼─────┼─────────────────────────┼────────────┼────────────┼──────────────┼──────
✓ │ 0   │ elrond-node-0.service   │ 0          │ 21345678   │ erd1...      │ 8080
✓ │ 1   │ elrond-node-1.service   │ 1          │ 21345679   │ erd1...      │ 8081
✗ │ 2   │ elrond-node-2.service   │ 2          │ -          │ -            │ 8082
✓ │ 3   │ elrond-node-3.service   │ metachain  │ 21345670   │ erd1...      │ 8083

health: 3 ok, 0 lagging, 1 failed, 0 unknown

Options

Flag	Default	Effect
`--watch`	off	Repaint until Ctrl+C.
`--interval <SECS>`	`5`	Refresh interval with `--watch`.
`--format table\|json`	`table`	Output shape.

mxnode status --watch --interval 3
mxnode --json status

Logs

mxnode logs --node 0 --follow

Options

Flag	Default	Effect
`--node N`	required	Node index. Repeatable for combined stream.
`--follow`	off	Tail new lines. Conflicts with `--save-archive`.
`--since <SPEC>`	none	Anything `journalctl --since` accepts (`1h`, `30min`, `2024-01-01`).
`--save-archive`	off	Tarball logs from every node into `<custom_home>/mx-chain-logs/`.

# Last hour, no follow
mxnode logs --node 0 --since 1h

# Combined tail of two nodes
mxnode logs --node 0 --node 1 --follow

# Archive everything
mxnode logs --save-archive

Dashboard

mxnode dashboard

Live TUI (built on ratatui). Press q to quit.

Options

Flag	Default	Effect
`--node N`	all	Limit to a subset (repeatable).
`--interval <ms>`	`1000`	REST poll cadence per node.
`--host <H>`	`127.0.0.1`	Override the host the dashboard talks to.
`--ws-logs`	off	Stream logs via the node’s `/log` WebSocket instead of tailing the file.

Metrics

mxnode metrics --port 9090

Prometheus scrape endpoint with per-node nonce, lag, and active state. Add it to your scrape config:

scrape_configs:
  - job_name: mxnode
    static_configs:
      - targets: ["host.example.com:9090"]

The metrics server runs in the foreground until Ctrl+C. Wrap it in your own systemd unit for production use.

What gets recorded

Action	Recorded where
Every state-changing op	`~/.local/state/mxnode/state.toml`
In-flight upgrades	`~/.local/state/mxnode/inflight.toml` (cleared on completion or auto-cleared on next run)
Migration audit log	`state.toml` `migrations.entries[]`

​Shortest commands

​Selectors

​Start / stop / restart

​Restart strategy

​Status

​Options

​Logs

​Options

​Dashboard

​Options

​Metrics

​What gets recorded

Shortest commands

Selectors

Start / stop / restart

Restart strategy

Status

Options

Logs

Options

Dashboard

Options

Metrics

What gets recorded