optional pod selector labels

charts/core-dump-handler/README.md

@@ -214,6 +214,7 @@ The agent pod has the following environment variables and these are all set by t
 * INTERVAL - The amount of time in milliseconds between each check of the core dump folder for files to upload.
 * SCHEDULE - A CRON formatted string [See cron library](https://github.com/mvniekerk/tokio-cron-scheduler#usage).
 * USE_INOTIFY - Set a listener for the coredump folder can be used in conjunction with SCHEDULE
+* COMP_POD_SELECTOR_LABEL - Optional selector label to filter pods that have core dump collection enabled. Default (empty) disables filter and enables collection for all. E.g. when selector label is set as "my.org/batch-workload" only pods that have a label named "my.org/batch-workload" (any value) will be enabled for core dump collection.
 
 ### Secrets
 
@@ -274,6 +275,7 @@ Composer
     namespace - the namespace the pod is associated with.
 
 * logLength: The amount of lines to take from the crashing pod. (Default 500)
+* podSelectorLabel: Enable composer only if pod has label matching the specified selector. (Default "" matches all pods)
 
 Daemonset
 * hostDirectory: Maps to the HOST_DIR environment variable (Default "/var/mnt/core-dump-handler")

charts/core-dump-handler/ci/inotify-manage-store.yaml

@@ -26,6 +26,7 @@ composer:
   # Double curlies are required otherwise helm trys to parse the string
   filenameTemplate: "{{uuid}}-dump-{{timestamp}}-{{hostname}}-{{exe_name}}-{{pid}}-{{signal}}"
   logLength: 500
+  podSelectorLabel: ""
 
 daemonset:
   name: "core-dump-handler"

charts/core-dump-handler/ci/interval-manage-store.yaml

@@ -26,6 +26,7 @@ composer:
   # Double curlies are required otherwise helm trys to parse the string
   filenameTemplate: "{{uuid}}-dump-{{timestamp}}-{{hostname}}-{{exe_name}}-{{pid}}-{{signal}}"
   logLength: 500
+  podSelectorLabel: ""
 
 daemonset:
   name: "core-dump-handler"

charts/core-dump-handler/ci/schedule-no-manage-store.yaml

@@ -26,6 +26,7 @@ composer:
   # Double curlies are required otherwise helm trys to parse the string
   filenameTemplate: "{{uuid}}-dump-{{timestamp}}-{{hostname}}-{{exe_name}}-{{pid}}-{{signal}}"
   logLength: 500
+  podSelectorLabel: ""
 
 daemonset:
   name: "core-dump-handler"

charts/core-dump-handler/ci/tolerations.yaml

@@ -27,6 +27,7 @@ composer:
   # Double curlies are required otherwise helm trys to parse the string
   filenameTemplate: "{{uuid}}-dump-{{timestamp}}-{{hostname}}-{{exe_name}}-{{pid}}-{{signal}}"
   logLength: 500
+  podSelectorLabel: ""
 
 daemonset:
   name: "core-dump-handler"

charts/core-dump-handler/templates/daemonset.yaml

@@ -41,6 +41,8 @@ spec:
             value: {{ .Values.composer.ignoreCrio | quote }}
           - name: COMP_CRIO_IMAGE_CMD
             value:  {{ .Values.composer.crioImageCmd }}
+          - name: COMP_POD_SELECTOR_LABEL
+            value:  {{ .Values.composer.podSelectorLabel }}
           - name: DEPLOY_CRIO_CONFIG
             value:  {{ .Values.daemonset.deployCrioConfig | quote }}
           - name: CRIO_ENDPOINT

charts/core-dump-handler/values.schema.json

@@ -112,6 +112,9 @@
                 },
                 "crioImageCmd": {
                     "type": "string"
+                },
+                "podSelectorLabel": {
+                    "type": "string"
                 }
             },
             "required": [

charts/core-dump-handler/values.yaml

@@ -26,6 +26,7 @@ composer:
   logLevel: "Warn"
   filenameTemplate: "{uuid}-dump-{timestamp}-{hostname}-{exe_name}-{pid}-{signal}"
   logLength: 500
+  podSelectorLabel: ""
 
 daemonset:
   name: "core-dump-handler"

core-dump-agent/src/main.rs

@@ -472,11 +472,12 @@ fn create_env_file(host_location: &str) -> Result<(), std::io::Error> {
         "{uuid}-dump-{timestamp}-{hostname}-{exe_name}-{pid}-{signal}".to_string()
     });
     let log_length = env::var("LOG_LENGTH").unwrap_or_else(|_| "500".to_string());
+    let pod_selector_label = env::var("COMP_POD_SELECTOR_LABEL").unwrap_or_default();
     info!("Creating {} file with LOG_LEVEL={}", destination, loglevel);
     let mut env_file = File::create(destination)?;
     let text = format!(
-        "LOG_LEVEL={}\nIGNORE_CRIO={}\nCRIO_IMAGE_CMD={}\nUSE_CRIO_CONF={}\nFILENAME_TEMPLATE={}\nLOG_LENGTH={}\n",
-        loglevel, ignore_crio, crio_image, use_crio_config, filename_template, log_length
+        "LOG_LEVEL={}\nIGNORE_CRIO={}\nCRIO_IMAGE_CMD={}\nUSE_CRIO_CONF={}\nFILENAME_TEMPLATE={}\nLOG_LENGTH={}\nPOD_SELECTOR_LABEL={}\n",
+        loglevel, ignore_crio, crio_image, use_crio_config, filename_template, log_length, pod_selector_label
     );
     info!("Writing composer .env \n{}", text);
     env_file.write_all(text.as_bytes())?;

core-dump-agent/tests/basic.rs

@@ -95,7 +95,7 @@ fn basic() -> Result<(), std::io::Error> {
         "FILENAME_TEMPLATE={uuid}-dump-{timestamp}-{hostname}-{exe_name}-{pid}-{signal}"
     ));
     assert!(env_content.contains("LOG_LENGTH=500"));
-    assert_eq!(env_content.lines().count(), 6);
+    assert_eq!(env_content.lines().count(), 7);
     //TODO: [No9] Test uploading of a corefile
     //TODO: [No9] Test remove option
     //TODO: [No9] Test sweep option

core-dump-composer/src/config.rs

@@ -18,6 +18,7 @@ pub struct CoreConfig {
     pub crictl_config_path: PathBuf,
     pub log_level: String,
     pub log_length: u32,
+    pub pod_selector_label: String,
     pub use_crio_config: bool,
     pub ignore_crio: bool,
     pub image_command: ImageCommand,
@@ -85,6 +86,7 @@ impl CoreConfig {
         let mut base_path = env::current_exe()?;
         base_path.pop();
 
+        let pod_selector_label = env::var("POD_SELECTOR_LABEL").unwrap_or_default();
         let log_level = env::var("LOG_LEVEL").unwrap_or_default();
         let ignore_crio = env::var("IGNORE_CRIO")
             .unwrap_or_else(|_| "false".to_string())
@@ -122,6 +124,7 @@ impl CoreConfig {
 
         Ok(CoreConfig {
             log_level,
+            pod_selector_label,
             ignore_crio,
             dot_env_path,
             image_command,

core-dump-composer/src/main.rs

@@ -58,6 +58,25 @@ fn main() -> Result<(), anyhow::Error> {
         }
     };
 
+    // match the label filter if there's one, and skip the whole process if it doesn't match
+    if !cc.pod_selector_label.is_empty() {
+        debug!(
+            "Pod selector specified. Will record only if pod has label {}",
+            &cc.pod_selector_label
+        );
+        let pod_labels = pod_object["labels"].as_object().unwrap();
+        // check if pod_labels has pod_selector_label
+        if pod_labels.get(&cc.pod_selector_label).is_none() {
+            info!(
+                "Skipping pod as it did not match selector label {}",
+                &cc.pod_selector_label
+            );
+            process::exit(0);
+        }
+    } else {
+        debug!("No pod selector specified, selecting all pods");
+    }
+
     let namespace = pod_object["metadata"]["namespace"]
         .as_str()
         .unwrap_or("unknown");