omq 0.17.0 → 0.17.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: ec5d4d1943efe37da6ad5493f80591b9578467e0c470fd4e2a577968580609cd
-  data.tar.gz: 8476b90cc637629874c94388cd91ae7a6f3ce3aaac410af1aa77c3fbd6470412
+  metadata.gz: f8a35ab8a212a111e78889494ac1eb18f780b6315eb3993eb83d04ae7ebdb222
+  data.tar.gz: 32727fe82814e3d36843dcbd82d66ecfbd9c26e5c36e7a173bdcf164dd649f75
 SHA512:
-  metadata.gz: 67b62142c8786b9594efdf5f62f0aef8b5329498a04faafec760305cb73d3de354b6f4dd7f5db3e416f4e69e683b1c60ad7920a879fb72c7942d602af2088715
-  data.tar.gz: 2731913ad5262a1d0038af1887b3c2bb167df719b650dd99d3d59812573180adb0e528eeaafba021823f12d7e8a38dbad92b8df07a9d6bac98a4534b71ae3cfc
+  metadata.gz: 1f78b8ff8a907b606ec376a7186a09b77b91317c91f3bdb1ff7ebd600d36d6e6fea4738b72dc4ddd3c611225df4a9c27fbd8418fe73dfcdf7a20985fdee5516b
+  data.tar.gz: 3811837748c18669dd6dc604c8253cfdb7b0e26a6c9eda3b61a4c711ac441f88b58c283701e6985880aa03d963338b53cc4271c3e456e6499498f9ce21037d85

data/CHANGELOG.md

@@ -1,5 +1,35 @@
 # Changelog
 
+## 0.17.2 — 2026-04-10
+
+### Fixed
+
+- **Reconnect after handshake failure.** When a peer RST'd a TCP
+  connection mid-ZMTP-handshake (e.g. `LINGER 0` close against an
+  in-flight connect), `ConnectionLifecycle#handshake!` called
+  `transition!(:closed)` directly, bypassing `tear_down!` and its
+  `maybe_reconnect` call. `spawn_connection`'s `ensure close!` then
+  saw the state already `:closed` and did nothing — the endpoint died
+  silently with no reconnect ever scheduled. Now the handshake rescue
+  goes through `tear_down!(reconnect: true)`, emitting `:disconnected`
+  and scheduling reconnect like any other connection loss.
+
+## 0.17.1 — 2026-04-10
+
+### Changed
+
+- **Reconnect sleeps are wall-clock quantized.** `Engine::Reconnect`
+  now sleeps until the next `delay`-sized grid tick instead of `delay`
+  from now (same math as `Async::Loop.quantized`). Multiple clients
+  reconnecting with the same interval wake up at the same instant,
+  collapsing staggered retries into aligned waves — easier to reason
+  about for observability and cache-warmup, and a server coming back
+  up sees one batch of accepts instead of a smear. Wall-clock (not
+  monotonic) on purpose: the grid has to line up across processes.
+  Anti-jitter by design. Exponential backoff still works: each
+  iteration quantizes to its own (growing) interval's grid, and
+  clients at the same backoff stage still align with each other.
+
 ## 0.17.0 — 2026-04-10
 
 ### Changed

data/lib/omq/engine/connection_lifecycle.rb

@@ -89,7 +89,11 @@ module OMQ
       rescue Protocol::ZMTP::Error, *CONNECTION_LOST => error
         @engine.emit_monitor_event(:handshake_failed, endpoint: @endpoint, detail: { error: error })
         conn&.close
-        transition!(:closed)
+        # Full tear-down with reconnect: without this, spawn_connection's
+        # ensure-block close! sees :closed and skips maybe_reconnect,
+        # leaving the endpoint dead. Race is exposed when a peer RSTs
+        # mid-handshake (e.g. LINGER 0 close against an in-flight connect).
+        tear_down!(reconnect: true)
         raise
       end
 

data/lib/omq/engine/reconnect.rb

@@ -42,7 +42,7 @@ module OMQ
         @engine.tasks << parent_task.async(transient: true, annotation: "reconnect #{@endpoint}") do
           loop do
             break if @engine.closed?
-            sleep delay if delay > 0
+            sleep quantized_wait(delay) if delay > 0
             break if @engine.closed?
             begin
               @engine.transport_for(@endpoint).connect(@endpoint, @engine)
@@ -60,6 +60,25 @@ module OMQ
 
       private
 
+      # Wall-clock quantized sleep: wait until the next +delay+-sized
+      # grid tick. Multiple clients reconnecting with the same interval
+      # wake up at the same instant, collapsing staggered retries into
+      # aligned waves. Same math as +Async::Loop.quantized+.
+      #
+      # Wall-clock (not monotonic) on purpose: the grid has to line up
+      # across processes, and monotonic clocks don't share an origin.
+      # Anti-jitter by design — if you want spread, don't call this.
+      #
+      # @param delay [Numeric] grid interval in seconds
+      # @param now [Float] wall-clock time in seconds (injectable for tests)
+      # @return [Float] seconds to sleep, always in (0, delay]
+      #
+      def quantized_wait(delay, now = Time.now.to_f)
+        wait = delay - (now % delay)
+        wait.positive? ? wait : delay
+      end
+
+
       def init_delay(delay)
         ri = @options.reconnect_interval
         if ri.is_a?(Range)

data/lib/omq/version.rb

@@ -1,5 +1,5 @@
 # frozen_string_literal: true
 
 module OMQ
-  VERSION = "0.17.0"
+  VERSION = "0.17.2"
 end

metadata

@@ -1,7 +1,7 @@
 --- !ruby/object:Gem::Specification
 name: omq
 version: !ruby/object:Gem::Version
-  version: 0.17.0
+  version: 0.17.2
 platform: ruby
 authors:
 - Patrik Wenger