Skip to content

Janzeer node 0.1.7 (mainnet)

Choose a tag to compare

@janzeerOrg janzeerOrg released this 17 Oct 06:38
· 5 commits to master since this release

Fifth mainnet release. Same network, same protocol: any 0.1.x node keeps working, upgrade when convenient.

  • Network janzeer, protocol 3.4.0, API 1.1.0
  • Genesis hash 3a716783ee58bc56335aaede1d489eeda22cd7442c358f085a83213028b7782c
  • Native executables, no Java needed
  • Versions 0.1.4 to 0.1.6 were internal builds used on the project's own validators and were never published; this release carries everything they had.

Fixed — syncing on slow links (Raspberry Pi, home Wi-Fi, VPN)

  • A 200-block sync page that took longer than 10 s to arrive was silently discarded on arrival, and the node asked a second and a third peer for the same page every 10 s: on a home link every page took a minute. Pages are no longer subject to the message expiry, a page still streaming is left alone, and a late duplicate is ignored (F-045).
  • A validator a block or two behind re-fetched the whole top page of the epoch (~330 KB) to get one block, which took a whole slot on a slow link and made it miss the next live block again. The serving peer now sends only the blocks above the requester's tip, so a catch-up carries one or two blocks and takes seconds (F-048; served by every 0.1.7 peer to any requester).
  • A sync page request lost in a connection hiccup could leave a node in PROCESSING for ever, silently (live blocks and votes on the channel looked like the page "still streaming"). The wait is now bounded: a request sent while the peer was disconnected is re-sent at the first expiry, no request is waited for longer than 2 minutes, and the node logs when it gives up on a peer (F-049).
  • A lagging validator proposing on its slot with a stale tip was logged as an ERROR ("Incorrect height in block") on every peer and paged operators running the alert script; it is now one WARN line on the proposer's peers (F-046).
  • A block-production slot firing during shutdown logged "Block creation failure" at ERROR; the production timer now stops with the node (F-047).
  • The bundled systemd unit declares SuccessExitStatus=143, so a clean systemctl stop is no longer logged as a failure.

Upgrade from 0.1.3

./update.sh (Linux, macOS, Raspberry Pi): downloads this bundle, verifies the minisign signature and the checksum, keeps data/, swaps the files and restarts the service (./update.sh --rollback restores 0.1.3). Or by hand: stop the node, replace the bundle folder, keep data/, start again. Arch: sudo pacman -U janzeer-node-bin-0.1.7-1-x86_64.pkg.tar.zst && sudo systemctl restart janzeer-node. Windows: stop the node, unpack the new zip over the old folder (keep data), run start.cmd.

SHA-256

beee979390188d5ac431692fa7e91d26d73b5a47cae1b44f4bf640ec9b6cf9e5  janzeer-node-0.1.7-mainnet-linux-x64.tar.gz
9d1804e1062f0e4dca6c1b7256e2fabe0b66e2003e3c50e857f40b5394160743  janzeer-node-bin-0.1.7-1-x86_64.pkg.tar.zst
6ef33616775a417bdf8931635052e614f87f41ca3da063bde067d1afa2ce49d8  janzeer-node-0.1.7-mainnet-linux-arm64.tar.gz
ac1bbdbab4001988dbc0b6faeaab0102295a583ecd1229e1ccd125c9dda25bf0  janzeer-node-0.1.7-mainnet-windows-x64.zip

SHA256SUMS is signed with the project's minisign key (key id 9F798583B345AE01, public key RWQBrkWzg4V5n5wUBa9GNPpa95Z7WFT2EY5QrabMv/DP+KG+gJsPIgdr): verify with minisign -Vm SHA256SUMS -P <key> and sha256sum -c SHA256SUMS before running.