Release v1.0.4 — Replication, Engine v3.3.32, SNI Policy & Metrics · SamNet-dev/MTProxyMax

What's New

Automatic config synchronization from a master server to one or more slaves via rsync+SSH on a configurable interval (default: 60s).

Synced files: secrets.conf, upstreams.conf, instances.conf, config.toml
Never synced: settings.conf, replication.conf (slave role always preserved)
Wizard: mtproxymax replication setup — interactive master/slave/standalone setup
Failover: mtproxymax replication promote — promote slave to master
Configurable SSH user (REPLICATION_SSH_USER), rsync --delete toggle, dependency checks
TUI: [r] Replication in main menu with full management interface
112 unit tests included

Configurable mask (permissive, default) or drop (strict) for TLS connections with non-matching SNI.

mtproxymax metrics — connections, upstream routing, per-user stats, ME pool status
mtproxymax metrics live [seconds] — auto-refresh dashboard

mtproxymax secret reset-traffic <label|all> — manually reset per-user cumulative traffic

Bounded hybrid routing loop — hard timeout on ME no-writer recovery
ArcSwap snapshots — lock-free concurrent reads, less contention
Parallel health checks — reduced latency during writer recovery
Refined quarantine — draining writers no longer needlessly quarantine healthy endpoints
New backpressure model — tiered base/high watermark
TLS fetcher redesign — adaptive profile cascade, per-target caching
Atomic per-user quotas — removed locking from hot path

Fix proxy auto-restarting after intentional stop (#49) — Docker restart policy and bot auto-recovery now respect a manual stop
Fix 'echo: write error: Broken pipe' on Alpine (#37) — replaced process-substitution FIFOs with here-strings
Fix menu requiring double input on Alpine (#38) — drain leftover escape-sequence bytes from multi-byte key presses
Fix 'Enter choice' prompt disappearing on Alpine (#48) — removed -s flag from input drain reads
Fix SNI rejection after engine upgrade (#40) — unknown_sni_action default changed to drop in v3.3.31+, now explicitly configurable
Fix traffic stats lost on restart (#44) — flush traffic before hot-reload; save existing cumulative even when Prometheus unreachable
Fix update lock leak (#43) — flock FD now released via trap RETURN; fixed false "already running" in same TUI session
Auto-clean old Docker images (#45) — old engine images pruned on every update
Re-exec after script update (#46) — TUI automatically restarts with the new version after update
Fix hot-reload — cp config in-place instead of mv to preserve Docker bind-mount inode

mtproxymax update