RpcNode Toolkit

Home/Versions

Agent channel · 0.4.214

Versions

What shipped in the host agent. Each release has its own page for search. Panel Servers compares each host to /install/TOOLKIT_VERSION. Feed: /install/VERSIONS.json.

Update an existing server from the panel. New hosts: install the agent.

  1. 0.4.214latest

    Check ports: no lsof hang on a busy host

    • Second node on a host with a live fullnode (Solana): check-ports no longer runs lsof after empty ss. One timed ss -lntp snapshot. Panel timeout is not shown as Ports busy.
  2. 0.4.213

    Check ports: no lsof hang on a busy host

    • Second node on a host with a live fullnode (Solana): check-ports no longer runs lsof after empty ss. One timed ss -lntp snapshot. Panel timeout is not shown as Ports busy.
  3. 0.4.212

    Release 0.4.212

    • Shipped via update-remote-agent.sh --publish.
  4. 0.4.211

    Host & node: free space per disk

    • Each disk tab shows free / total (largest FS on that disk) plus used %. LVM OS disk maps via /sys slaves.
  5. 0.4.210

    Release 0.4.210

    • Shipped via update-remote-agent.sh --publish.
  6. 0.4.209

    Host & node: per-disk IOPS tabs

    • metrics.json current.disks[] + history.disks[] — each whole physical disk has its own read/write IOPS, MB/s and %util.
    • Host & node shows a tab per disk (All = sum IOPS / hottest %util). Solana ledger vs accounts no longer collapse into one number.
  7. 0.4.208

    Release 0.4.208

    • Shipped via update-remote-agent.sh --publish.
  8. 0.4.207

    Host & node: disk IOPS / util

    • metrics.json now includes host disk_read_iops, disk_write_iops, disk_read_mb_s, disk_write_mb_s, disk_util_pct (hottest whole disk from /proc/diskstats — same signal as iostat).
    • Host & node shows Read / Write IOPS, MB/s and %util. Node unit IOPS from cgroup io.stat when IOAccounting is on.
  9. 0.4.206

    Remove: mask units so Restart=always cannot respawn

    • Panel remove now disable+mask+SIGTERM --kill-who=all before ACK. Restart=always / watchdog cannot bring leaf agents or the node back.
    • TON: oneshot RemainAfterExit (ton-<env>.service) is not 'still running'. Truth is validator-engine / mytoncore. Stock validator.service is masked, not deleted.
  10. 0.4.205

    Install: host disk insights (md / LVM / raw NVMe)

    • GET /api/v1/host/disks returns insights + unused disks: software RAID is a bottleneck, LVM / is the OS disk, raw NVMe is where node data should go.
    • Install wizard shows that summary for every network. JBOD recommend prefers raw NVMe over md /data.
  11. 0.4.204ton

    TON: catch-up stalled when oos does not drop

    • seqno>0 + ~27h behind + 0.1% is not 'won't start' — validator is up.
    • If local seqno rises but out_of_sync_sec grows: catch-up slower than tip (serializer / I/O). UI says that, not 0.1% lag closed.
  12. 0.4.203

    Release 0.4.203

    • Shipped via update-remote-agent.sh --publish.
  13. 0.4.202

    Release 0.4.202

    • Shipped via update-remote-agent.sh --publish.
  14. 0.4.201solana

    Solana: reuse Agave binary, do not rebuild the workspace

    • Install looks for agave-validator in /opt/solana/<env>/bin and /var/tmp/rpcnode-agave-src-*/{bin,target/release} and copies it if cargo already produced the file.
    • Source build is cargo build --release --bin agave-validator only. cargo-install-all.sh is not used (it compiles the whole workspace: reqwest, rustls, …).
    • Retry does not rm -rf an existing clone.
    • Identity uses /opt/solana/<env>/bin/solana-keygen (was empty optPath → keygen not found after a successful cargo build).
  15. 0.4.200

    Release 0.4.200

    • Shipped via update-remote-agent.sh --publish.
  16. 0.4.199solana

    Client download URLs in host audit log

    • Install writes every catalog/artifact URL to /var/log/rpcnode.log (action download + provision): manifest, /clients/→/install/clients/ rewrite, GET ok/FAIL.
    • Every hostLog line includes the agent version: [api-agent 0.4.199]. Solana Agave from CDN /install/clients/… (GitHub fallback). 0.4.198 on the host had no these lines.
    • agent.sh banner + already-installed menu show installed vs CDN channel version immediately.
  17. 0.4.197xrplton

    XRPL first-ledger recycle + TON dump apply

    • XRPL 0.4.196 skipped NuDB wipe while peers/proposers>0, but PrepareDatadirHeal still stopped xrpld every collect tick (default 2s). Skip stop/wipe while .first-ledger-wait is younger than 12 min. Stop xrpld only when NuDB will actually rotate.
    • TON: dump-end systemctl restart sets NRestarts=1 — that is not a crash loop (was forcing sticky 1G cache).
    • TON getstats seqno is masterchainblock (-1,wc,SEQ), not masterchainblocknumber. Prefer getstats over MyTonCtrl so the panel does not freeze at seqno=0 / dump 99% during catch-up.
  18. 0.4.196xrpl

    XRPL canonical UNL — first validated ledger

    • Stock validators-example.txt ships [validator_list_threshold] 0 → xrpld exits 255 (Invalid publisher key: 0).
    • Install and heal write canonical UNL (vl.ripple.com + unl.xrplf.org). Agent does not wipe NuDB while peers/proposers>0 waiting for validated_ledger.
  19. 0.4.191

    Client update steps in host audit log

    • Apply wrote only UI state «starting client update» — /var/log/rpcnode.log stayed empty.
    • Each step now hostLog: accepted, catalog, RPC sleep, soft-stop, install, start, done/error. Same for Restart / Stop.
  20. 0.4.190panel 0.1.33

    Soft-stop fullnode from the node page

    • Stop next to Restart: same graceful stop (CLI / ExecStop / SIGTERM), unit stays down until Restart.
    • Public Go RPC sleeps (503) while stopped. Chain data is not wiped.
  21. 0.4.188tronbscbitcoinethereum

    Client update apply for all catalog networks

    • Check already worked everywhere; Apply was hard-coded to TRON (and briefly BSC).
    • Apply is now the same pipeline on every leaf: soft-stop (CLI/RPC then systemctl), download the catalog artifact (jar / bin / tarball), start the same units.
    • Apt / docker_extract / empty artifact_url stay an honest error — there is nothing to replace.
  22. 0.4.187panel 0.1.32ton

    TON getVersion + install CDN host

    • Admin/node version check was status=unsupported — toncenter HTTP API has no version method.
    • Public Go RPC now answers version / getVersion / getNodeVersion with the same client_version as the panel (validator-engine --version).
    • Bare POST / is rewritten to /api/v2/jsonRPC so the public base URL works as a TON JSON-RPC endpoint.
    • Update no longer fetches rpcnode.dev/install (404). Channel is toolkit.rpcnode.dev; old env/unit URLs are rewritten.
  23. 0.4.186ton

    TON dump apply — stop killing validator at 99%

    • After a MyTonCtrl dump, seqno stayed 0 and the bar sat at 99% «applying state» for a day (testnet and mainnet).
    • MyTonCtrl rewrites validator ExecStart; heal no longer SIGKILL-recycles while apply is running.
    • After OOM or a restart, celldb cache is capped to 1G and MemoryMax=85%.
    • Sync shows restart / apply OOM instead of a fake «applying» line.
  24. 0.4.184panel 0.1.31cardanoxrpl

    Cardano Mithril snapshot + XRPL history picker

    • Cardano Snapshot step downloads official Mithril latest (include-ancillary), mithril-client 2630.0.
    • Honest snapshot % from N/7 files and MiB in the log.
    • XRPL install picker: stock / 1 day / 2 weeks / full history (default weeks).
  25. 0.4.183panel 0.1.30tron

    TRON mainnet snapshot flavors

    • Install picker: standard / internal_tx / balance_history — matching official LevelDB mirrors and HOCON.
    • Choice is persisted on the panel and on the host; re-add from awaiting_ports clears it.
  26. 0.4.179xrpl

    XRPL provision compile

    • ensureXRPLDInstalled takes the environment so panel Docker builds again.
  27. 0.4.178xrpl

    XRPL client pin 3.3.0

    • catalog.json pins xrpld 3.3.0. Empty version + unpinned apt was installing 3.2.0.
  28. 0.4.177xrpl

    XRPL first ledger + hub peers

    • seq=0 with overlay flap and 0–2 peers now pins official hubs in ips_fixed.
    • LoadManager stall recycles at medium node size.
  29. 0.4.176xrpl

    XRPL dead server_stop is not a start fault

    • no response from server / ExecMainCode=1 during ExecStop on a dead xrpld is ignored; the unit is started.
  30. 0.4.174panel 0.1.28ton

    TON OOM after dump is not Synced

    • seqno=0 after dump is apply/OOM, not tip — no 99.9% / 0.1% lag-closed bounce.
    • Cap --celldb-cache-size from RAM; strip preload-all / in-memory / fast-state-serializer.
  31. 0.4.163panel 0.1.25ethereum

    Ethereum tip after sync

    • Public tip shows blocks / headers after sync. eth_syncing=false has no highestBlock.
  32. 0.4.155xrpl

    XRPL history peers

    • peers_max=100, official hubs, ips_fixed s2.ripple.com. Heal on start.
    • No official snapshot — full history is still a long peer backfill.
  33. 0.4.154panel 0.1.18xrplstellarbitcoindogeltcdashbchzcash

    Synced means tip and history proof

    • Live tip alone is not Synced. XRPL complete_ledgers, Stellar oldestLedger, Bitcoin-family not pruned.
    • UI and panel require verification_pct ≥ 99.9.
  34. 0.4.153xrpl

    XRPL false Synced

    • server_state=full on a short complete_ledgers window is live tip, not history.
    • Synced only when complete_ledgers reaches genesis (32570) and the tip.
  35. 0.4.152panel 0.1.17xrpl

    XRPL Sync bar from complete_ledgers

    • verification_pct from complete_ledgers high / tip. No official tarball (~39 TiB via peers).
  36. 0.4.149panel 0.1.16

    Port holder + Kill

    • Check ports shows the program holding a catalog port. Kill (with confirm) goes through the tip agent.
  37. 0.4.137panel 0.1.14

    Check ports does not parse ufw

    • Reach is a panel dial to the Server public IP. The agent never runs ufw allow / deny.
  38. 0.4.136panel 0.1.13

    Outbound reach on public / agent / p2p

    • Check ports: panel dials the Server public IP while the tip holds a nonce listener. reach = reachable or filtered.
  39. 0.4.117

    Host audit log

    • /var/log/rpcnode.log is discrete hostLog ops only (update, provision, start, snapshot, errors). Agent stdout stays per-unit.
  40. 0.4.116tron

    TRON Java 8

    • ensureJava8 checks java -version. Missing or wrong Java → apt openjdk-8-jre-headless, else Temurin 8.
    • Unit ExecStart is pinned to Java 8 so a host default of 17 does not crash GreatVoyage.
  41. 0.4.115ethereumrobinhoodoptimismbasearb

    Sepolia L1 RPC for Orbit / OP-stack

    • Parent-chain default is publicnode ethereum-sepolia. Mainnet L1 stays on the host ethereum port.
    • Override with RPCNODE_L1_RPC_URL.
  42. 0.4.112optimismbitcoin

    Optimism sepolia + Bitcoin Core 28.1

    • Optimism sepolia provision (was a mainnet-only stub). Bitcoin Core default 28.1 for testnet4.
  43. 0.4.111basebscarb

    Base sepolia JWT, BSC cache, Arb sepolia

    • Write JWT before waiting on authrpc. Sepolia parent-chain uses sepolia L1.
    • BSC --cache capped to ~25% RAM. Arb sepolia provision added.
  44. 0.4.110

    Host OS deps on provision and start

    • The agent apt-installs what the network needs. TRON Java 8 is no longer a manual SSH step.