HTND v2.18.2 Release Notes
This release is optional, but highly recommended and is focussed on correctness,stability, and security fixes.
Action required: secure your RPC listener
HTND now binds RPC to 127.0.0.1 by default instead of 0.0.0.0
If you require RPC to be reachable from things other than loopback please add --rpclisten=0.0.0.0 to restore the old behaviour
--saferpc is also now safer, and disables some other dangerous commands.
Other changes
-
Safer shutdowns: HTND now waits for active RPC handlers and queued consensus work before closing the database, preventing shutdown-time database-closed crashes. If it cannot shut down safely within the timeout, it exits without closing the database; - - Pebble recovers through its write-ahead log on next start.
-
More reliable pruning-point updates: pruning-point UTXO-set updates are deferred until after a block is committed and relayed, avoiding long consensus-lock pauses in the critical relay path. Failed updates remain pending and are retried.
-
P2P hardening: peers that send unsolicited blocks are disconnected instead of allowing blocks to accumulate in memory.
-
Shutdown race fix: a connection whose router starts during shutdown no longer causes a panic.
-
Various security fixes that reduce the nodes vulnerability to being crashed by remote network traffic.
2.18.2 has been through a set of automated testing suites on testnet, and has successfully passed them. As we continue to develop this automated test capability more details will be routinely supplied with future updates.
Running a node in 6GB RAM
Experimental.
#!/bin/bash
export GOMEMLIMIT=3500MiB
export GOGC=50
export HTND_PEBBLE_CACHE_MB=1024
export HTND_BLOCK_CACHE_MB=256
./HTND --utxoindex --saferpc --rpclisten=0.0.0.0 -a htn.foztor.net -a 219.88.72.130
It is entirely possible the node will still OOM if subject to heavy mining/compounding load, but this configuration has been successfully running for over 72 hours with no OOM.