1.3.0 (#88)

* added ability to overcommit

Author: NorseGaud

README.md

@@ -104,6 +104,7 @@ pid_file_dir: /tmp/
 # global_database_database: 0
 # global_private_key: /Users/{YOUR USER HERE}/.private-key.pem # If you use the same key for all your plugins, you can set it here.
 # global_token: github_pat_XXX # If you use the same token for all your plugins, you can set it here.
+# global_skip_cpu_and_memory_resource_checks: true # Optional; skip VM CPU/RAM admission checks for all handler plugins (allows overcommit).
 # plugins_path: ~/.config/anklet/plugins/ # This sets the location where scripts used by plugins are stored; we don't recommend changing this.
 plugins:
 ```
@@ -136,6 +137,7 @@ It is also possible to use ENVs for several of the items in the config. They ove
 | ANKLET_GLOBAL_TOKEN | GitHub token for all plugins (ex: github_pat_XXX) |
 | ANKLET_GLOBAL_RECEIVER_SECRET | Secret to use for receiver plugin (ex: "my-secret") |
 | ANKLET_GLOBAL_TEMPLATE_DISK_BUFFER | Disk buffer (how much disk space to leave free on the host) percentage for templates (ex: 10.0 for 10%) |
+| ANKLET_GLOBAL_SKIP_CPU_AND_MEMORY_RESOURCE_CHECKS | Skip VM CPU/RAM admission checks for all handler plugins (ex: true) |
 
 You can also set or override plugin settings per plugin using envs based on the plugin name:
 

VERSION

@@ -1 +1 @@
-1.2.2
+1.3.0

internal/config/config.go

@@ -40,6 +40,7 @@ type Config struct {
 	GlobalDatabaseTLSInsecure        bool     `yaml:"global_database_tls_insecure"`
 	GlobalReceiverSecret             string   `yaml:"global_receiver_secret"`
 	GlobalTemplateDiskBuffer         float64  `yaml:"global_template_disk_buffer"` // Global disk buffer percentage (e.g., 10.0 for 10%)
+	GlobalSkipCPUAndMemoryResourceChecks bool `yaml:"global_skip_cpu_and_memory_resource_checks"` // Global override to skip CPU/RAM resource admission checks
 }
 
 type Log struct {
@@ -98,6 +99,7 @@ type Plugin struct {
 	RegistrationTimeoutSeconds int      `yaml:"registration_timeout_seconds"`
 	TemplateDiskBuffer         float64  `yaml:"template_disk_buffer"` // Plugin-specific disk buffer percentage (e.g., 10.0 for 10%)
 	JobRetryAttempts           int      `yaml:"job_retry_attempts"`   // Maximum number of retry attempts for failed jobs
+	SkipCPUAndMemoryResourceChecks bool `yaml:"skip_cpu_and_memory_resource_checks"` // Skip CPU/RAM resource admission checks for this plugin
 }
 
 // GetQueueOwner returns QueueName if set, otherwise returns Owner.
@@ -250,6 +252,10 @@ func LoadInEnvs(config Config) (Config, error) {
 		}
 		config.GlobalTemplateDiskBuffer = buffer
 	}
+	envGlobalSkipCPUAndMemoryResourceChecks := os.Getenv("ANKLET_GLOBAL_SKIP_CPU_AND_MEMORY_RESOURCE_CHECKS")
+	if envGlobalSkipCPUAndMemoryResourceChecks != "" {
+		config.GlobalSkipCPUAndMemoryResourceChecks = envGlobalSkipCPUAndMemoryResourceChecks == "true"
+	}
 
 	envJobRetryAttempts := os.Getenv("ANKLET_JOB_RETRY_ATTEMPTS")
 	if envJobRetryAttempts != "" {

internal/config/config_test.go

@@ -257,10 +257,11 @@ func TestApplyPluginEnvOverrides(t *testing.T) {
 
 func TestApplyPluginEnvOverrides_BoolFields(t *testing.T) {
 	tests := []struct {
-		name         string
-		envVars      map[string]string
-		pluginPrefix string
-		wantSkipPull bool
+		name                                 string
+		envVars                              map[string]string
+		pluginPrefix                         string
+		wantSkipPull                         bool
+		wantSkipCPUAndMemoryResourceChecks   bool
 	}{
 		{
 			name:         "skip_pull true",
@@ -274,6 +275,12 @@ func TestApplyPluginEnvOverrides_BoolFields(t *testing.T) {
 			pluginPrefix: "TEST_PLUGIN",
 			wantSkipPull: false,
 		},
+		{
+			name:       "skip_cpu_and_memory_resource_checks true",
+			envVars:    map[string]string{"TEST_PLUGIN_SKIP_CPU_AND_MEMORY_RESOURCE_CHECKS": "true"},
+			pluginPrefix: "TEST_PLUGIN",
+			wantSkipCPUAndMemoryResourceChecks: true,
+		},
 	}
 
 	for _, tt := range tests {
@@ -292,6 +299,13 @@ func TestApplyPluginEnvOverrides_BoolFields(t *testing.T) {
 			if plugin.SkipPull != tt.wantSkipPull {
 				t.Errorf("SkipPull = %v, want %v", plugin.SkipPull, tt.wantSkipPull)
 			}
+			if plugin.SkipCPUAndMemoryResourceChecks != tt.wantSkipCPUAndMemoryResourceChecks {
+				t.Errorf(
+					"SkipCPUAndMemoryResourceChecks = %v, want %v",
+					plugin.SkipCPUAndMemoryResourceChecks,
+					tt.wantSkipCPUAndMemoryResourceChecks,
+				)
+			}
 		})
 	}
 }

main.go

@@ -152,6 +152,9 @@ func main() {
 			if plugin.TemplateDiskBuffer == 0 {
 				loadedConfig.Plugins[index].TemplateDiskBuffer = 10.0
 			}
+			if loadedConfig.GlobalSkipCPUAndMemoryResourceChecks {
+				loadedConfig.Plugins[index].SkipCPUAndMemoryResourceChecks = true
+			}
 		}
 	}
 

plugins/handlers/github/README.md

@@ -38,6 +38,7 @@ plugins:
     # registration_timeout_seconds: 80 # Optional; defaults to 80 seconds.
     # template_disk_buffer: 10.0 # Optional; defaults to 10.0%. How much disk space to leave free on the host for templates.
     # job_retry_attempts: 10  # Override the default of 5
+    # skip_cpu_and_memory_resource_checks: true # Optional; allows CPU/RAM overcommit by skipping VM admission checks on this host.
     #database:
     #  enabled: true
     #  url: localhost
@@ -57,6 +58,7 @@ plugins:
 - If you are attempting to register runners for an entire organization, do NOT set `repo` and make sure your Github App has `Self-hosted runners` > `Read and write` permissions.
 - If your Organization level runner is registered and your public repo jobs are not picking it up even though the labels are a perfect match, make sure the Runner groups (likely `Default`) has `Allow public repositories`.
 - There are times when github will not register the runner for some reason. `registration_timeout_seconds` is available to set the custom seconds to wait before considering the runner registration failed (and retry on a new VM).
+- `skip_cpu_and_memory_resource_checks` (or global `global_skip_cpu_and_memory_resource_checks`) is intended for overcommit fleets. It allows multiple VMs to start even when host CPU/RAM capacity checks would normally block them, which can improve throughput in bursty workloads but may increase contention and job runtime variance under sustained load.
 
 ---