cuboid 0.5.1 → 0.5.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: '090c49b680b5086f8f5a8c4abef0dacdb8bc0e68b7f2f0d37d9592bc82a34c66'
-  data.tar.gz: bebcf857aad1397f85a2b0aa430fbd2862803cb03bfa03ca4a91453aaf31b229
+  metadata.gz: 600130025a98049b1378e4598204ec17e2816095779b2b66f88ce7167a1a647f
+  data.tar.gz: b84d8da4ee6844556c043884c3f74f3e7c66992f940bf72d52f0a18784630cfd
 SHA512:
-  metadata.gz: 316eafd64fdd9db5db029de782fbac6cdf8534ccfcc85c26103ad68ee1665b2d4d365ebf6c86e08a9c93805cee2eecf829e5247ac4ef5525d3d2d5ba3b6b6477
-  data.tar.gz: 4329d042508b387abc6e802be493ef55c9f9d3824bbd5aa51f21d8b0b006bea7a46a0749618be66ebbedecc586e420884d84127293686cf1d4aa88d139a46be7
+  metadata.gz: 293059b05e93cd107840a6d7985749eaee19d979113ccb58ea98ee1f5738278ce795bef2ef6eea313a7d516bbf6564815195a6a60af20154175feb44170c755e
+  data.tar.gz: 4d9555ac474bae2a4f7a6cdc341bddff3406e4b23b71fa685da3d738dc454d5b8f59018996b99963391b26e4b7171c28bc0ee18f23a03ec5da67dd06bfcaebce

data/lib/cuboid/rpc/server/scheduler.rb

@@ -45,6 +45,13 @@ class Scheduler
 
     TICK_CONSUME = 0.1
 
+    # Number of CONSECUTIVE RPC errors a running Instance is allowed to
+    # accumulate before it's given up on and marked {#failed}. A single
+    # transient blip (the Instance momentarily unreachable on one ping, an
+    # engine reconnect, etc.) shouldn't fail an otherwise-healthy scan; any
+    # clean response in between resets the count.
+    RPC_ERROR_TOLERANCE = 5
+
     def initialize
         @options = Options.instance
 
@@ -62,9 +69,11 @@ class Scheduler
         @id_to_priority = {}
         @by_priority    = {}
 
-        @running   = {}
-        @completed = {}
-        @failed    = {}
+        @running           = {}
+        @completed         = {}
+        @failed            = {}
+        # Per-Instance consecutive RPC-error counters (see RPC_ERROR_TOLERANCE).
+        @rpc_error_counts  = Hash.new( 0 )
 
         set_handlers( @server )
         trap_interrupts { Thread.new { shutdown } }
@@ -335,17 +344,37 @@ class Scheduler
                 handle_rpc_error( id, busy )
                 block.call
             elsif busy
+                # Clean response — the Instance is reachable and working, so
+                # reset its consecutive-error tolerance.
+                @rpc_error_counts.delete( id )
                 print_debug "[#{id}] Busy."
                 block.call
             else
+                # Don't reset here: this leads to the report grab, whose own
+                # transient errors must be allowed to accumulate toward the
+                # tolerance rather than being cleared every ping.
                 get_report_and_shutdown( id, client, &block )
             end
         end
     end
 
     def handle_rpc_error( id, error )
-        print_error "[#{id}] Failed: [#{error.class}] #{error.to_s}"
+        @rpc_error_counts[id] += 1
+        count = @rpc_error_counts[id]
+
+        # Tolerate transient errors — only give up after RPC_ERROR_TOLERANCE
+        # of them in a row. The Instance stays in @running and gets re-checked
+        # on the next ping; a clean response resets the count.
+        if count < RPC_ERROR_TOLERANCE
+            print_error "[#{id}] RPC error #{count}/#{RPC_ERROR_TOLERANCE} " \
+                        "(tolerating): [#{error.class}] #{error}"
+            return
+        end
+
+        print_error "[#{id}] Failed after #{count} consecutive RPC errors: " \
+                    "[#{error.class}] #{error}"
 
+        @rpc_error_counts.delete( id )
         @failed[id] = {
             error:       error.class.to_s,
             description: error.to_s
@@ -371,6 +400,7 @@ class Scheduler
             client.shutdown do
                 print_status "[#{id}] Completed."
 
+                @rpc_error_counts.delete( id )
                 @running.delete( id ).close
                 @completed[id] = path
 

data/lib/version

@@ -1 +1 @@
-0.5.1
+0.5.2

metadata

@@ -1,14 +1,14 @@
 --- !ruby/object:Gem::Specification
 name: cuboid
 version: !ruby/object:Gem::Version
-  version: 0.5.1
+  version: 0.5.2
 platform: ruby
 authors:
 - Tasos Laskos
 autorequire:
 bindir: bin
 cert_chain: []
-date: 2026-06-10 00:00:00.000000000 Z
+date: 2026-06-12 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: awesome_print