Tron

Documentation Index

Fetch the complete documentation index at: https://docs.novacula.io/llms.txt

Use this file to discover all available pages before exploring further.

Tron nodes run as a single java-tron process per node. The chain adapter installs a JRE matched to the host architecture, renders config.conf (HOCON), and starts java-tron with a network-specific JVM heap size. An optional snapshot-fetch init container can seed the LevelDB data directory from a community-operated mirror.

​

Networks and resources

JVM heap is set per network and exposed as a managed value — the adapter passes -Xms<heap> -Xmx<heap> to the JVM. To tune it, edit heapSize in the adapter’s network config; it is not a per-node user override.

​

Ports

​

Node types

Full only. For full-history use cases, set node.openHistoryQueryWhenLiteFN: false (the default) and accept the disk budget.

​

Init steps

On first reconcile per node:

1. JRE install — the right JRE archive for the host architecture is downloaded into the node’s data directory. This step runs once per (host, version) pair.
2. Optional snapshot fetch — if a SNAPSHOT_MIRRORS URL is set for the network (mainnet today; Nile is sync-from-genesis), an init container walks the mirror’s nginx autoindex to pick the freshest /backup<YYYYMMDD>/ directory, downloads the LevelDB archive, verifies SHA-256, and unpacks into the data directory. There is no stable /latest/ symlink upstream — the walker picks the date-named directory on every fresh init.
3. Config render — config.conf is written from a HOCON template merged with whitelisted user overrides.

​

Config overrides

Overrides go into config.conf (HOCON). The schema declares:

• Managed paths — networking, data dir, ports, mirror endpoints. Rejected at validate time.
• Known keys — a curated TRON_USER_ALLOWED_KEYS list (see packages/adapter/src/tron/apply-config.ts) covering tunables like:
  • node.openHistoryQueryWhenLiteFN (boolean)
  • node.fastSyncSupport and other *Args flags (boolean)
  • Numeric tuning under node.* and storage.*

Free-form keys not on the managed list are accepted, but the typical workflow is to stick to the known list. The adapter uses HOCON-aware merging (appendHoconAssignments) to splice user overrides into the rendered file.

​

JVM tuning

The biggest knob for java-tron performance is heap size. Defaults assume modern mainnet block sizes; if you’re running on substantially more RAM, bump heapSize in your adapter config and let it grow proportionally. Setting heap > 70% of node RAM is asking for OOM kills under load — leave headroom for the JVM’s off-heap and the kernel page cache.

​

Metrics

Scraped from the java-tron Prometheus exporter on :9527. Surfaced:

• Current block height, solidified block height.
• Peer count.
• JVM heap usage, GC stats.
• Disk usage.

See Node monitoring.

​

Common patterns

• Lite full node — set node.openHistoryQueryWhenLiteFN: true; node serves recent history only and uses much less disk.
• Snapshot-bootstrapped mainnet — defaults — adapter init pulls the latest /backup<date>/ from the mainnet mirror, cutting initial sync time substantially. Nile sync is genesis-only (no published mirror).
• Behind a proxy — expose :8090 (HTTP) or :50051 (gRPC) through your own reverse proxy; the adapter doesn’t manage TLS termination.