Skip to main content

Zcash Foundation

Back to Blog

November 19, 2026

Foundation News

Zebra 7.2.0-beta.1 Release

Zebra 7.2.0-beta.1: Faster Mining and Sync, Plus Block Pruning, Ready for Miners to Test

We are pleased to announce Zebra 7.2.0-beta.1, a beta for miners, mining pools and the operators who run nodes for them. It brings most of the mining improvements we first tested in Zebra Ándale together with new performance work: faster verification and initial sync, faster block propagation, and block templates ready sooner after each new block. It also adds opt in block pruning, which shrank one Mainnet database from about 282 GB to about 15 GB. If you run Zcash mining infrastructure, please run it beside 7.1.0, watch the metrics, and tell us what you find: your reports shape the stable 7.2.0 release. Production nodes should stay on Zebra 7.1.0, but results from test nodes are welcome too, especially initial sync times.

For Miners and Pools: New Block Templates Sooner, Faster Propagation, Less Disk

Since NU7, Zcash targets a block every 25 seconds (ZIP 218), a third of the old 75, so the time between a new block and new work for your miners, and the time before your own block reaches the network, is a bigger share of every block. This beta shortens both: Zebra hands out a new block template sooner, announces blocks built from its templates while it is still verifying them, and passes blocks between upgraded nodes faster. With pruning, a mining node can also run on much less disk.

A New Block Template Sooner After Each Block

Zebra now serves each template as soon as it is built and checks it in the background (#12507), and refreshes templates when the mempool changes (#11374). Once a new block’s proof of work and header checks pass, Zebra can hand out coinbase only work on top of it before its transaction checks finish; that work expires after 1 second (#11669). The internal miner starts on new work immediately instead of sleeping for up to 20 seconds (#11692).

Your Blocks Get Out Sooner

When you submit a block built from a template Zebra already checked, Zebra announces it to peers while it is still being verified, instead of after it is committed (#12507). It also skips validating the same transactions again (#11658) and checks the block’s place in the chain while the block is still being verified, so submitblock returns sooner (#12542).

Blocks also travel faster between nodes running this beta. Zebra announces a block it received as soon as it commits it, instead of after up to 7 seconds, and announces mempool transactions every 2 seconds instead of every 7 (#11693). Compact block relay, on by default, rebuilds new blocks from the mempool instead of downloading them whole, and lets them pass between supporting peers once their proof of work and transaction commitments check out (#11932). Transactions a peer announced but can’t send are requested from other peers that announced them after 200 milliseconds, instead of up to 73 seconds later (#12532). Compact block relay uses a service bit from a draft ZIP; network.compact_block_relay = false turns it off.

Less Disk With Pruning

Set state.prune_blocks_older_than and Zebra keeps at least 10,000 recent blocks plus everything it needs to validate, mine and submit blocks (#12092), which suits a mining node or a pool backend that doesn’t need old blocks for payouts or accounting. In one Mainnet measurement in the pruning guide, a pruned database took about 15 GB, against about 282 GB for an archive; leave extra room for initial sync and compaction. Pruning is off by default and doesn’t work with a zcashd sidecar. Pruned state lives in a separate state-pruned/v29 directory, and moving an existing archive there is a manual step with the node stopped, so back it up first and follow the guide. Pruned nodes advertise NODE_NETWORK_LIMITED. zcashd doesn’t download blocks from them and Zebra 7.1.0 only downloads the blocks they announce, so keep some Zebra peers connected, for example with network.persistent_peers, so the blocks you find spread quickly (#12537).

Faster Verification and Initial Sync

Batched proofs and signatures now go to verification after 5 milliseconds without new input (#11673), shielded transactions are started before transparent ones (#11805), and the cryptography libraries behind Orchard, Ironwood and Sapling are faster revisions (#12492). Note commitment trees start updating while a block’s other checks run (#12172), tip commits no longer copy the indexes of recent blocks (#12508), and transparent scripts verified in the mempool aren’t verified again in a block (#11939).

Initial sync up to the checkpoints (block hashes that ship with Zebra) now does more in parallel: preparing block data before the database writer needs it (#11694), reading large blocks in parallel, storing transactions without encoding them again (#12419), and updating note commitment trees on a dedicated thread pool (#12420).

Run It Beside Your Current Node

Some of these changes come from cryptography library revisions pinned by commit until they are published, so we want them on real mining workloads before they reach every node:

  1. Run the beta beside your 7.1.0 node, and point a test miner or a small share of pool work at it.
  2. Get work through getblocktemplate long polling, which also delivers the early work on new blocks. notify.zmqpubhashblock (source builds only) fires only once a block is committed, so use it alongside long polling, not instead of it.
  3. Compare, from your miner or pool software, how soon new work arrives after each block, and whether your solved blocks are accepted and relayed. On the beta, rpc.getblocktemplate.first_template.duration_seconds and rpc.submitblock.outcomes.total show the node’s side.
  4. Tell us what you find on GitHub.

Breaking Changes

Regtest and Custom Testnet Operators

Zebra now checks custom network configurations at startup instead of failing later: a Regtest node rejects overrides of its fixed values such as network_magic (#12049), Regtest configs with too few funding stream addresses need the missing addresses or extend_funding_stream_addresses_as_required = true (#12048), and lockbox recipients must be P2SH or P2PKH addresses (#12360). The release notes list every check.

Wallet, RPC and lightwalletd Services

As specified in ZIP 2009, this beta requires the original ZIP 317 fees until Mainnet reaches height 3,590,000, expected around 15:30 UTC on November 19, 2026. Until then, transactions that pay the reduced fee of 1,000 zatoshis per logical action, which 7.1.0 accepts, are not relayed or mined. From that height on, and already on public Testnet, it accepts the same fees as 7.1.0. (#12365)

Dashboards and Alerts

rpc.request.duration_seconds and zcash.net.peer.handshake.duration_seconds are now Prometheus histograms, and several metric labels changed; the release notes list them. The dashboards and alert rules that ship with Zebra are already updated.

New Features

Opt In Fast Tree Sync

With state.fast_tree_sync on, initial sync uses note commitment tree roots supplied by peers, checked against Zebra’s checkpoints, instead of computing every tree itself (#12427). It is a one way change that Zebra 7.1.0 and tools reading the database directly can’t use, so keep a backup from before you enable it. Leave it off on nodes that serve wallets: at fast synced heights, z_gettreestate and the lightwalletd tree state methods return an “unavailable” error. Its peer messages are a draft pending a ZIP.

For Mining Pools

getmininginfo reports the network difficulty and getpeerinfo each peer’s starting height (#12062), and submitblock and JSON-RPC 1.0 batches now pass the startup checks of pools such as s-nomp (#11668). notify.zmqpubhashblock publishes hashblock notifications in builds from source with --features zmq; the release binaries and Docker images refuse to start with it set (#12075).

preciousblock, Persistent Peers and Metrics

The new preciousblock RPC tells Zebra which of two equal work chain tips to prefer (#12114). network.persistent_peers keeps connections to peers you choose, and when network.external_addr isn’t set, Zebra advertises the public IP address its outbound peers agree on (#12459). Zebra also exports standard Prometheus process metrics (#12437) and timings for each phase of block verification and template building (#12479).

Bug Fixes

  • getaddressutxos is about 50 times faster for addresses with a long, fully spent history, such as the funding stream addresses (#12390).
  • Initial sync no longer stalls on pruned Zakura peers (Zakura 1.8.8 and later) (#12521).
  • Supplied Grafana panels and alert rules that queried metrics Zebra doesn’t export now work, and the errors dashboard is removed.

The release notes list the rest, including a mempool panic and clean shutdown on SIGINT and SIGTERM.

Other Security Improvements

The example Docker Compose files no longer publish Zebra’s RPC or metrics ports to the host, and the observability stack needs an admin password to start (#12039). If you use the example mining stack, rename MINER_DATA to EXTRA_COINBASE_DATA in your .env file (#12040).

State Database Format

On its first start, the beta upgrades a 7.0.0, 7.1.0 or Zebra Ándale state from format 29.0.0 to 29.2.0 in place, in about a minute on Mainnet, with no resync. A 7.2.0-beta.0 state is already on 29.2.0.

Upgrading

Get the beta from GitHub or Docker Hub as zfnd/zebra:7.2.0-beta.1, also tagged edge; zfnd/zebra:latest stays on 7.1.0. Configs from 7.1.0 or 7.2.0-beta.0 on Mainnet or public Testnet load unchanged, but if you tested Zebra Ándale, remove network.experimental_early_block_relay first: this beta rejects it. Upgrade any pruned 7.2.0-beta.0 node, since beta.0 tells older peers it keeps the full history. The release notes list every change, and the Zebra book for this beta has been rewritten.

Switching back to 7.1.0. A node that never enabled pruning or fast tree sync can go back to 7.1.0 without resyncing. First remove the new network.compact_block_relay, network.persistent_peers, state.prune_blocks_older_than, state.fast_tree_sync, sync.pin_parallel_cpu_threads and notify.zmqpubhashblock fields from your config, because 7.1.0 rejects fields it doesn’t know. A database that has been pruned or fast synced can’t go back, so keep a backup from before you enable either option. Don’t switch a pruned node back to 7.2.0-beta.0.

The beta has the same end of support as 7.1.0: it starts warning around December 3, 2026, and stops itself around December 10, 2026. If something looks wrong, please open an issue on GitHub.

Thank You to Our Contributors

This release was made possible by the work and reviews of @arcayne, @arya2, @conradoplg, @gustavovalverde, @jvff, @oxarbitrage, and @upbqdn. Thank you for your continued contributions to Zebra.

Zebra is the Zcash Foundation’s independent, Rust-based implementation of the Zcash protocol. Learn more at github.com/ZcashFoundation/zebra.