Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion client/config/config.cpp
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@

const NodeVersion NODE_VERSION = 532;
// the community release (4.2.532-community.N); also announced to peers in host_data bits 32-47
const uint8_t MINOR_NODE_VERSION = 14;
const uint8_t MINOR_NODE_VERSION = 15;

const std::string BLOCK_NAME_PARAMS = "params";
const std::string BLOCK_NAME_HOST_INPUT = "host_input";
Expand Down
13 changes: 13 additions & 0 deletions docs/CHANGES.md
Original file line number Diff line number Diff line change
Expand Up @@ -25,6 +25,9 @@ lines under the new release.
starts in minutes from several independent sources. See [snapshots.md](snapshots.md).
- **Windows and macOS.** Packages for Windows (x64) and macOS (Apple Silicon), with a mainnet check
that the three systems agree on the state.
- **Full nodes from a snapshot.** A full node on LMDB or RocksDB starts from the snapshot in minutes
and fetches the older blocks from its neighbours in the background (backward sync), each checked
against the chain; consensus keeps priority. See [storage.md](storage.md).
- **Wallet state digest.** A digest of all wallet state, logged every 1000 blocks, lets different
nodes and builds compare their state.
- **Continuous integration.**
Expand All @@ -37,6 +40,10 @@ lines under the new release.

## Not released yet

Nothing yet.

## community.15 — 2026-11-01

| PR | Change |
|---|---|
| #119 | ci: the release checks the Windows and macOS packages on mainnet before publishing them (a node of each system restores the published snapshot, syncs to the head, logs the same state digests as Linux and stops cleanly), and the same check runs weekly on the latest published release |
Expand Down Expand Up @@ -69,6 +76,12 @@ lines under the new release.
| #146 | csnode: the backfill turn budget depends on the node's role: 50 ms while it is a confidant (`backfill_turn_budget_ms`), 90 ms otherwise (`backfill_idle_turn_budget_ms`); on mainnet a node outside the trusted set backfills at about 450 blocks/s instead of about 250 |
| #147 | csnode: while the node is a confidant, the backfill storage flush (median 358 ms, up to 2.4 s on mainnet, on the processor thread) waits until the node is out of the trusted set, 10 minutes at most; blocks not flushed are fetched again after a crash |
| #148 | csnode: the periodic backfill flush runs on a thread of its own instead of the processor thread (on mainnet it took up to 5.6 s, with the storing of blocks waiting); one at a time, the final flush waits for the one under way, and the node stops the thread before its storage closes |
| — | release: the version announced to peers (`MINOR_NODE_VERSION`) is 15; the packages ship the contract executor community-2 (build 1555) |

Backward sync (#122–#125, #129, #130, #133, #136, #139, #140, #145–#148). A full node on LMDB or
RocksDB starts from the snapshot and fetches the whole history in the background; on mainnet about
400 blocks/s while the node is not a confidant, about five days for the 186 million blocks.
Security (#142–#144): see the rows above; they were reported privately to the Credits team first.

## community.14 — 2026-10-27

Expand Down
28 changes: 27 additions & 1 deletion docs/FEATURES.md
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@ consensus with them. This page lists what it adds; each item links the pull requ
The full list of changes is in [CHANGES.md](CHANGES.md), the operational details in
[differences.md](differences.md).

Everything listed here is in release 4.2.532-community.14 or earlier.
Everything listed here is in release 4.2.532-community.15 or earlier.

## Storage

Expand All @@ -29,6 +29,12 @@ Everything listed here is in release 4.2.532-community.14 or earlier.
rebuilt from the stored blocks instead of re-reading the chain from block 0.
[#53](https://github.com/EnzinoBB/node/pull/53), [#54](https://github.com/EnzinoBB/node/pull/54),
[#56](https://github.com/EnzinoBB/node/pull/56)
- **A full node in minutes, the history in the background.** A full node on lmdb or rocksdb starts
from the snapshot and then fetches the older blocks from its neighbours down to block 0, checking
each against the hash chain and its signatures; the API serves every block and history as it
arrives. Consensus comes first: the backfill gets a bounded share of the processor thread, less
while the node is a confidant. A pruned node switched to full fetches what it pruned the same way.
[#122](https://github.com/EnzinoBB/node/pull/122), [#123](https://github.com/EnzinoBB/node/pull/123), [#125](https://github.com/EnzinoBB/node/pull/125), [#129](https://github.com/EnzinoBB/node/pull/129), [#130](https://github.com/EnzinoBB/node/pull/130), [#133](https://github.com/EnzinoBB/node/pull/133), [#140](https://github.com/EnzinoBB/node/pull/140), [#145](https://github.com/EnzinoBB/node/pull/145), [#146](https://github.com/EnzinoBB/node/pull/146), [#147](https://github.com/EnzinoBB/node/pull/147)

## Starting a node in minutes: snapshots

Expand Down Expand Up @@ -89,6 +95,14 @@ Everything listed here is in release 4.2.532-community.14 or earlier.
work on nodes that quick-started or pruned the blocks involved.
[#20](https://github.com/EnzinoBB/node/pull/20), [#46](https://github.com/EnzinoBB/node/pull/46),
[#74](https://github.com/EnzinoBB/node/pull/74)
- **Tokens in the API.** A token cache that follows every token standard (`[api] tokens_cache=true`),
contract results after a quick start, and contract calls that no method matches refused before
their fee is paid; a token template and a Python client in `samples/`.
[#126](https://github.com/EnzinoBB/node/pull/126), [#128](https://github.com/EnzinoBB/node/pull/128), [#132](https://github.com/EnzinoBB/node/pull/132), [#134](https://github.com/EnzinoBB/node/pull/134), [#135](https://github.com/EnzinoBB/node/pull/135)
- **Executor community-2** (build 1555): tokens of the v2 standard are reported as tokens, and
contract code that runs past its time limit, or whose caller has gone, is stopped. Contract results
are the same as with the official build 1518.
[#137](https://github.com/EnzinoBB/node/pull/137)

## Consensus robustness

Expand All @@ -102,6 +116,10 @@ Everything listed here is in release 4.2.532-community.14 or earlier.
delegations are restored and reverted correctly.
[#8](https://github.com/EnzinoBB/node/pull/8), [#9](https://github.com/EnzinoBB/node/pull/9),
[#11](https://github.com/EnzinoBB/node/pull/11), [#12](https://github.com/EnzinoBB/node/pull/12)
- **Forward sync under special order 35.** A syncing node no longer asks a neighbour for blocks that
neighbour would refuse as a duplicate, and asks another one instead: a node syncing among mostly
pruned peers could otherwise wait for ever for its first blocks.
[#136](https://github.com/EnzinoBB/node/pull/136)

## Crash and safety fixes

Expand All @@ -113,6 +131,11 @@ the storage engines (a full disk, a failed write) that hung or stopped the node.
[#57](https://github.com/EnzinoBB/node/pull/57), [#59](https://github.com/EnzinoBB/node/pull/59),
[#63](https://github.com/EnzinoBB/node/pull/63)

Since community.15: packet parsers bounded by the bytes received, a network stop request that no
longer stops the node, a utility message obeyed only with the blockchain key's signature, size
limits in the p2p layer (a fork of the p2p library), and two races in the cache databases.
[#121](https://github.com/EnzinoBB/node/pull/121), [#124](https://github.com/EnzinoBB/node/pull/124), [#142](https://github.com/EnzinoBB/node/pull/142), [#143](https://github.com/EnzinoBB/node/pull/143), [#144](https://github.com/EnzinoBB/node/pull/144)

## Packages and Docker

- **Release packages** for Ubuntu 20.04 and later (one package: `config.ini` full, `config.pruned.ini`
Expand All @@ -132,6 +155,9 @@ the storage engines (a full disk, a failed write) that hung or stopped the node.
[#109](https://github.com/EnzinoBB/node/pull/109)
- **Drop-in upgrade.** An official node's folder upgrades to the community node and back.
[#43](https://github.com/EnzinoBB/node/pull/43), [#44](https://github.com/EnzinoBB/node/pull/44)
- **Guided set-up** in every package (`setup.sh`, `setup.bat`): pruned or full, database, how the
first start gets the chain, snapshot sharing and the contract executor.
[#131](https://github.com/EnzinoBB/node/pull/131)

## Tested on every change

Expand Down
8 changes: 5 additions & 3 deletions docs/storage.md
Original file line number Diff line number Diff line change
Expand Up @@ -83,9 +83,11 @@ transaction lookups and wallet histories grow downwards while the node runs, wit
restart.

- **By default a full node started from a snapshot fetches the whole history to disk.** On mainnet
that is about 186 million blocks: about two days at the default `backfill_rate=1000` (longer if
the peers pace the requests), and the disk of a full node, about 400 GB with lmdb or berkeleydb,
about 130 GB with rocksdb. Set `backfill=off` to keep only the blocks of the snapshot.
that is about 186 million blocks: about five days (measured on mainnet: 300 to 450 blocks/s while
the node is not a confidant, less while it is, since consensus comes first; see
`backfill_idle_turn_budget_ms`), and the disk of a full node, about 400 GB with lmdb or
berkeleydb, about 130 GB with rocksdb (the recent blocks take about 1.3 KB each in rocksdb: keep
room to spare). Set `backfill=off` to keep only the blocks of the snapshot.
- Requests go out only while the node is in sync (not syncing forward, not in a bootstrap round,
not fetching a snapshot); a trusted node keeps fetching, at most `backfill_store_per_turn` blocks
per turn of the thread that also runs consensus.
Expand Down
2 changes: 1 addition & 1 deletion packaging/docker/README.md
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,7 @@ docker compose up -d && docker compose logs -f

A pruned node (the default) starts from a snapshot instead of syncing from block 0 for days. On its first start, with `CREDITS_SNAPSHOT=github` (the default), the container runs `node --download-snapshot`: it downloads the latest snapshot published on GitHub, checks every file against its signed manifest and restores it. When that fails, the node looks for a snapshot among its p2p neighbours (the community nodes that serve one) and, after `fetch_wait` (10 minutes) without one, syncs from block 0. `CREDITS_SNAPSHOT=p2p` skips the GitHub download, `off` syncs from block 0. The node checks a restored snapshot's state digest at its first start and stops if it differs.

A full node on lmdb or rocksdb (`CREDITS_STORAGE=full`, `CREDITS_DB_BACKEND=lmdb` or `rocksdb`) starts from the snapshot the same way. Then it fetches the older blocks from its neighbours in the background, down to block 0 (backward sync). It takes part in the network meanwhile, and the API serves each block and history as it arrives. On mainnet that is about 186 million blocks: about two days, and the disk of a full node, about 400 GB with lmdb or 130 GB with rocksdb. `CREDITS_BACKFILL=off` keeps only the snapshot's blocks. A full node on berkeleydb (the full default) syncs from block 0, as the official node does: a berkeleydb database cannot start above block 0. See `docs/storage.md`, "Backward sync".
A full node on lmdb or rocksdb (`CREDITS_STORAGE=full`, `CREDITS_DB_BACKEND=lmdb` or `rocksdb`) starts from the snapshot the same way. Then it fetches the older blocks from its neighbours in the background, down to block 0 (backward sync). It takes part in the network meanwhile, and the API serves each block and history as it arrives. On mainnet that is about 186 million blocks: about five days, and the disk of a full node, about 400 GB with lmdb or 130 GB with rocksdb. `CREDITS_BACKFILL=off` keeps only the snapshot's blocks. A full node on berkeleydb (the full default) syncs from block 0, as the official node does: a berkeleydb database cannot start above block 0. See `docs/storage.md`, "Backward sync".

The node also serves its own snapshots to its neighbours (`CREDITS_SERVE_SNAPSHOT=on`, the default; at most 2 MB/s, half while it is a confidant). See `docs/snapshots.md`.

Expand Down
2 changes: 1 addition & 1 deletion packaging/make-package.sh
Original file line number Diff line number Diff line change
Expand Up @@ -135,7 +135,7 @@ Start: cd csnode && $start (full storage, as th
A pruned node can start from the published snapshot: $start --config-file config.pruned.ini --download-snapshot
A full node can too, on lmdb or rocksdb: set db_backend=lmdb (or rocksdb) in the [storage] section
of config.ini, run $start --download-snapshot, then start it. It fetches the older blocks from its
neighbours in the background down to block 0 (backward sync: about two days and 400 GB with lmdb,
neighbours in the background down to block 0 (backward sync: about five days and 400 GB with lmdb,
130 GB with rocksdb, on mainnet; [storage] backfill=off keeps only the snapshot's blocks). A
pruned node switched to full (prune_keep_blocks=0) fetches the blocks it pruned the same way.
https://github.com/EnzinoBB/node/blob/master/docs/storage.md#a-full-node-from-a-snapshot
Expand Down
Loading