Feat/multi git sync

CHANGELOG.md

@@ -4,6 +4,7 @@
 
 ### Added
 
+- Add git-sync support for multiple repositories ([#729]).
 - Add support for airflow 3.1.6 ([#742]).
 - Add operator versioning ([#725]).
   - GitSync considered for v1alpha1 and v1alpha2
@@ -21,6 +22,7 @@
 - Prevent unnecessary Pod restarts when initially creating an AirflowCluster.
   This is achieved by applying the StatefulSet after all ConfigMaps and Secrets that it mounts ([#734]).
 
+[#729]: https://github.com/stackabletech/airflow-operator/pull/729
 [#725]: https://github.com/stackabletech/airflow-operator/pull/725
 [#726]: https://github.com/stackabletech/airflow-operator/pull/726
 [#727]: https://github.com/stackabletech/airflow-operator/pull/727

docs/modules/airflow/pages/usage-guide/mounting-dags.adoc

@@ -42,7 +42,7 @@ For multiple DAGs, it is easier to expose them via `gitsync`, as shown below.
 
 == Via `git-sync`
 
-{git-sync}[git-sync] is a command that pulls a git repository into a local directory and is supplied as a sidecar container for use within Kubernetes.
+{git-sync}[git-sync] is a command that pulls a git repository into a local directory and is supplied as a sidecar container for use within Kubernetes. {git-sync}[git-sync] folders will be provided at `/stackable/app/allDAGs/current-i`, with i in {0,1,..,n-1}.
 The Stackable Airflow images already ship with git-sync included, and the operator takes care of calling the tool and mounting volumes, so that only the repository and synchronization details are required:
 
 .git-sync usage example: https
@@ -82,3 +82,24 @@ include::example$example-airflow-gitsync-ssh.yaml[]
 
 NOTE: git-sync can be used with DAGs that make use of Python modules, as Python is configured to use the git-sync target folder as the "root" location when looking for referenced files.
 See the xref:usage-guide/applying-custom-resources.adoc[] example for more details.
+
+=== Multiple repositories via git-sync
+
+If you want to access multiple branches of a repository or load dags from multiple repositories, you can extend the list shown above.
+
+The default mount-path for git-sync resources is `/stackable/app/allDAGs/current-{i}` where i corresponds to:
+
+[source,yaml]
+----
+    dagsGitSync:
+      - repo: ssh://git@github.com/stackable-airflow/dags.git # <-- current-0
+        branch: test/git-sync-0
+        credentials:
+          sshPrivateKeySecretName: git-sync-ssh
+      - repo: ssh://git@github.com/stackable-airflow/dags.git # <-- current-1
+        branch: test/git-sync-1
+        credentials:
+          sshPrivateKeySecretName: git-sync-ssh
+----
+
+NOTE: Using DAGs from ConfigMaps will require you to either use celeryExecutors and mount them under `/stackable/app/allDAGs/<name>` or use the kubernetesExecutor and mount them somewhere else, changing the PYTHONPATH (if using submodules) and `AIRFLOW\__CORE__DAGS_FOLDER` accordingly.

rust/operator-binary/src/airflow_controller.rs

@@ -83,9 +83,9 @@ use crate::{
     config::{self, PYTHON_IMPORTS},
     controller_commons::{self, CONFIG_VOLUME_NAME, LOG_CONFIG_VOLUME_NAME, LOG_VOLUME_NAME},
     crd::{
-        self, AIRFLOW_CONFIG_FILENAME, APP_NAME, AirflowClusterStatus, AirflowConfig,
-        AirflowConfigOptions, AirflowExecutor, AirflowRole, CONFIG_PATH, Container, ExecutorConfig,
-        ExecutorConfigFragment, HTTP_PORT, HTTP_PORT_NAME, LISTENER_VOLUME_DIR,
+        self, AIRFLOW_CONFIG_FILENAME, AIRFLOW_DAGS_FOLDER, APP_NAME, AirflowClusterStatus,
+        AirflowConfig, AirflowConfigOptions, AirflowExecutor, AirflowRole, CONFIG_PATH, Container,
+        ExecutorConfig, ExecutorConfigFragment, HTTP_PORT, HTTP_PORT_NAME, LISTENER_VOLUME_DIR,
         LISTENER_VOLUME_NAME, LOG_CONFIG_DIR, METRICS_PORT, METRICS_PORT_NAME, OPERATOR_NAME,
         STACKABLE_LOG_DIR, TEMPLATE_LOCATION, TEMPLATE_NAME, TEMPLATE_VOLUME_NAME,
         authentication::{
@@ -113,6 +113,7 @@ use crate::{
     },
 };
 
+pub const AIRFLOW_DAGS_INIT_VOLUME: &str = "all-dags-volume";
 pub const AIRFLOW_CONTROLLER_NAME: &str = "airflowcluster";
 pub const DOCKER_IMAGE_BASE_NAME: &str = "airflow";
 pub const AIRFLOW_FULL_CONTROLLER_NAME: &str =
@@ -1019,7 +1020,6 @@ fn build_server_rolegroup_statefulset(
             executor,
             authentication_config,
             authorization_config,
-            git_sync_resources,
             resolved_product_image,
         )
         .context(BuildStatefulsetEnvVarsSnafu)?,
@@ -1299,7 +1299,6 @@ fn build_executor_template_config_map(
             airflow,
             env_overrides,
             merged_executor_config,
-            git_sync_resources,
             resolved_product_image,
         ))
         .add_volume_mounts(airflow.volume_mounts())
@@ -1319,6 +1318,41 @@ fn build_executor_template_config_map(
         true,
     )?;
 
+    /*
+    Multi-gitsync requires an init-container which copies contents from
+    gitsync volumes into one empty volume to be mounted into
+    the airflow image.
+    */
+    let mut dags_init_container =
+        ContainerBuilder::new("multi-git-sync-init").context(InvalidContainerNameSnafu)?;
+    dags_init_container
+        .image_from_product_image(resolved_product_image)
+        .args(airflow.get_kubernetes_executer_multi_gitsync_commands())
+        .command(vec![
+            "/bin/bash".to_string(),
+            "-x".to_string(),
+            "-euo".to_string(),
+            "pipefail".to_string(),
+            "-c".to_string(),
+        ])
+        .add_volume_mounts(git_sync_resources.git_content_volume_mounts.clone())
+        .context(AddVolumeMountSnafu)?
+        .add_volume_mount(
+            AIRFLOW_DAGS_INIT_VOLUME.to_owned(),
+            AIRFLOW_DAGS_FOLDER.to_owned(),
+        )
+        .context(AddVolumeMountSnafu)?;
+
+    airflow_container
+        .add_volume_mount(
+            AIRFLOW_DAGS_INIT_VOLUME.to_owned(),
+            AIRFLOW_DAGS_FOLDER.to_owned(),
+        )
+        .context(AddVolumeMountSnafu)?;
+
+    pb.add_init_container(dags_init_container.build());
+    pb.add_volume(VolumeBuilder::new(AIRFLOW_DAGS_INIT_VOLUME.to_owned()).build())
+        .context(AddVolumeSnafu)?;
     pb.add_container(airflow_container.build());
     pb.add_volumes(airflow.volumes().clone())
         .context(AddVolumeSnafu)?;

rust/operator-binary/src/crd/mod.rs

@@ -71,6 +71,7 @@ pub const STACKABLE_LOG_DIR: &str = "/stackable/log";
 pub const LOG_CONFIG_DIR: &str = "/stackable/app/log_config";
 pub const AIRFLOW_HOME: &str = "/stackable/airflow";
 pub const AIRFLOW_CONFIG_FILENAME: &str = "webserver_config.py";
+pub const AIRFLOW_DAGS_FOLDER: &str = "/stackable/app/allDAGs";
 
 pub const TEMPLATE_VOLUME_NAME: &str = "airflow-executor-pod-template";
 pub const TEMPLATE_LOCATION: &str = "/templates";
@@ -439,6 +440,42 @@ impl v1alpha2::AirflowCluster {
         fragment::validate(conf_rolegroup).context(FragmentValidationFailureSnafu)
     }
 
+    // Softlink from each single folder git-{i} into {AIRFLOW_DAGS_FOLDER}/.
+    pub fn get_multi_gitsync_commands(&self) -> Vec<String> {
+        let mut symlinks = Vec::<String>::new();
+        for (i, _) in self.spec.cluster_config.dags_git_sync.iter().enumerate() {
+            symlinks.push(
+                format!("ln -s /stackable/app/git-{i}/current {AIRFLOW_DAGS_FOLDER}/current-{i}")
+                    .to_string(),
+            )
+        }
+        symlinks
+    }
+
+    // kubernetesExecuter needs to copy into an empty volume since git-{0}/current
+    // is a softlink and thus can't be shared via a volume mount (volume remains empty).
+    pub fn get_kubernetes_executer_multi_gitsync_commands(&self) -> Vec<String> {
+        let mut cp_commands = Vec::<String>::new();
+        for (i, _) in self.spec.cluster_config.dags_git_sync.iter().enumerate() {
+            cp_commands.push(
+                format!("cp -r /stackable/app/git-{i}/current/ {AIRFLOW_DAGS_FOLDER}/current-{i}")
+                    .to_string(),
+            );
+        }
+        // init-container seems to only accept one command line.
+        vec![cp_commands.join(" && ")]
+    }
+
+    // PYTHONPATH contains folder-name provided in CRD.
+    pub fn get_gitsync_absolute_paths(&self) -> Vec<String> {
+        let mut python_path = Vec::<String>::new();
+        for (i, git_sync) in self.spec.cluster_config.dags_git_sync.iter().enumerate() {
+            let folder = &git_sync.git_folder.display();
+            python_path.push(format!("{AIRFLOW_DAGS_FOLDER}/current-{i}/{folder}").to_string())
+        }
+        python_path
+    }
+
     /// Retrieve and merge resource configs for the executor template
     pub fn merged_executor_config(
         &self,
@@ -600,10 +637,13 @@ impl AirflowRole {
             format!(
                 "cp -RL {CONFIG_PATH}/{AIRFLOW_CONFIG_FILENAME} {AIRFLOW_HOME}/{AIRFLOW_CONFIG_FILENAME}"
             ),
+            // Adding cm as dags within the same AIRFLOW_DAGS_FOLDER may lead to problems, thus checking if exists.
+            format!("mkdir -p {AIRFLOW_DAGS_FOLDER}"),
             // graceful shutdown part
             COMMON_BASH_TRAP_FUNCTIONS.to_string(),
             remove_vector_shutdown_file_command(STACKABLE_LOG_DIR),
         ];
+        command.extend(airflow.get_multi_gitsync_commands());
 
         if resolved_product_image.product_version.starts_with("3.") {
             // Start-up commands have changed in 3.x.
@@ -663,6 +703,8 @@ impl AirflowRole {
                     container_debug_command(),
                     "airflow triggerer &".to_string(),
                 ]),
+                // KubernetesExecutor intentionally not covered, it requires command
+                // generated by airflow and written into a pod template.
                 AirflowRole::Worker => command.extend(vec![
                     "prepare_signal_handlers".to_string(),
                     container_debug_command(),
@@ -715,6 +757,8 @@ impl AirflowRole {
                     container_debug_command(),
                     "airflow triggerer &".to_string(),
                 ]),
+                // KubernetesExecutor intentionally not covered, it requires command
+                // generated by airflow and written into a pod template.
                 AirflowRole::Worker => command.extend(vec![
                     "prepare_signal_handlers".to_string(),
                     container_debug_command(),

rust/operator-binary/src/env_vars.rs

@@ -6,17 +6,15 @@ use std::{
 use product_config::types::PropertyNameKind;
 use snafu::Snafu;
 use stackable_operator::{
-    commons::product_image_selection::ResolvedProductImage,
-    crd::{authentication::oidc, git_sync},
-    k8s_openapi::api::core::v1::EnvVar,
-    kube::ResourceExt,
+    commons::product_image_selection::ResolvedProductImage, crd::authentication::oidc,
+    k8s_openapi::api::core::v1::EnvVar, kube::ResourceExt,
     product_logging::framework::create_vector_shutdown_file_command,
 };
 
 use crate::{
     crd::{
-        AirflowExecutor, AirflowRole, ExecutorConfig, LOG_CONFIG_DIR, STACKABLE_LOG_DIR,
-        TEMPLATE_LOCATION, TEMPLATE_NAME,
+        AIRFLOW_DAGS_FOLDER, AirflowExecutor, AirflowRole, ExecutorConfig, LOG_CONFIG_DIR,
+        STACKABLE_LOG_DIR, TEMPLATE_LOCATION, TEMPLATE_NAME,
         authentication::{
             AirflowAuthenticationClassResolved, AirflowClientAuthenticationDetailsResolved,
         },
@@ -80,14 +78,13 @@ pub fn build_airflow_statefulset_envs(
     executor: &AirflowExecutor,
     auth_config: &AirflowClientAuthenticationDetailsResolved,
     authorization_config: &AirflowAuthorizationResolved,
-    git_sync_resources: &git_sync::v1alpha2::GitSyncResources,
     resolved_product_image: &ResolvedProductImage,
 ) -> Result<Vec<EnvVar>, Error> {
     let mut env: BTreeMap<String, EnvVar> = BTreeMap::new();
     let secret = airflow.spec.cluster_config.credentials_secret.as_str();
     let internal_secret_name = airflow.shared_internal_secret_secret_name();
 
-    env.extend(static_envs(git_sync_resources));
+    env.extend(static_envs(airflow));
 
     // environment variables
     let env_vars = rolegroup_config.get(&PropertyNameKind::Env);
@@ -154,12 +151,11 @@ pub fn build_airflow_statefulset_envs(
         );
     }
 
-    let dags_folder = get_dags_folder(git_sync_resources);
     env.insert(
         AIRFLOW_CORE_DAGS_FOLDER.into(),
         EnvVar {
             name: AIRFLOW_CORE_DAGS_FOLDER.into(),
-            value: Some(dags_folder),
+            value: Some(AIRFLOW_DAGS_FOLDER.to_owned()),
             ..Default::default()
         },
     );
@@ -288,42 +284,32 @@ pub fn build_airflow_statefulset_envs(
     Ok(transform_map_to_vec(env))
 }
 
-pub fn get_dags_folder(git_sync_resources: &git_sync::v1alpha2::GitSyncResources) -> String {
-    let git_sync_count = git_sync_resources.git_content_folders.len();
-    if git_sync_count > 1 {
-        tracing::warn!(
-            "There are {git_sync_count} git-sync entries: Only the first one will be considered.",
-        );
+// contains absolute path to git-sync folder and log config.
+fn construct_python_path(airflow: &v1alpha2::AirflowCluster) -> String {
+    let mut python_path = LOG_CONFIG_DIR.to_string();
+    let symlinks = airflow.get_gitsync_absolute_paths();
+    // append `:` only of there are git-sync entries.
+    if !symlinks.is_empty() {
+        python_path.push(':');
+        python_path.push_str(symlinks.join(":").as_str());
     }
 
-    // If DAG provisioning via git-sync is not configured, set a default value
-    // so that PYTHONPATH can refer to it. N.B. nested variables need to be
-    // resolved, so that /stackable/airflow is used instead of $AIRFLOW_HOME.
-    // see https://airflow.apache.org/docs/apache-airflow/stable/configurations-ref.html#dags-folder
-    git_sync_resources
-        .git_content_folders_as_string()
-        .first()
-        .cloned()
-        .unwrap_or("/stackable/airflow/dags".to_string())
+    python_path
 }
 
 // This set of environment variables is a standard set that is not dependent on any
 // conditional logic and should be applied to the statefulset or the executor template config map.
-fn static_envs(
-    git_sync_resources: &git_sync::v1alpha2::GitSyncResources,
-) -> BTreeMap<String, EnvVar> {
+fn static_envs(airflow: &v1alpha2::AirflowCluster) -> BTreeMap<String, EnvVar> {
     let mut env: BTreeMap<String, EnvVar> = BTreeMap::new();
 
-    let dags_folder = get_dags_folder(git_sync_resources);
-
     env.insert(
         PYTHONPATH.into(),
         EnvVar {
             // PYTHONPATH must be extended to include the dags folder so that dag
             // dependencies can be found: this must be the actual path and not a variable.
             // Also include the airflow site-packages by default (for airflow and kubernetes classes etc.)
             name: PYTHONPATH.into(),
-            value: Some(format!("{LOG_CONFIG_DIR}:{dags_folder}")),
+            value: Some(construct_python_path(airflow)),
             ..Default::default()
         },
     );
@@ -372,7 +358,6 @@ pub fn build_airflow_template_envs(
     airflow: &v1alpha2::AirflowCluster,
     env_overrides: &HashMap<String, String>,
     config: &ExecutorConfig,
-    git_sync_resources: &git_sync::v1alpha2::GitSyncResources,
     resolved_product_image: &ResolvedProductImage,
 ) -> Vec<EnvVar> {
     let mut env: BTreeMap<String, EnvVar> = BTreeMap::new();
@@ -407,17 +392,16 @@ pub fn build_airflow_template_envs(
 
     // the config map also requires the dag-folder location as this will be passed on
     // to the pods started by airflow.
-    let dags_folder = get_dags_folder(git_sync_resources);
     env.insert(
         AIRFLOW_CORE_DAGS_FOLDER.into(),
         EnvVar {
             name: AIRFLOW_CORE_DAGS_FOLDER.into(),
-            value: Some(dags_folder),
+            value: Some(AIRFLOW_DAGS_FOLDER.to_owned()),
             ..Default::default()
         },
     );
 
-    env.extend(static_envs(git_sync_resources));
+    env.extend(static_envs(airflow));
 
     add_version_specific_env_vars(
         airflow,

tests/templates/kuttl/mount-dags-gitsync/30-install-airflow-cluster.yaml.j2

@@ -77,17 +77,26 @@ spec:
     dagsGitSync:
 {% if test_scenario['values']['access'] == 'ssh' %}
       - repo: ssh://git@github.com/stackable-airflow/dags.git
+        branch: test/git-sync-0
+        credentials:
+          sshPrivateKeySecretName: git-sync-ssh
+      - repo: ssh://git@github.com/stackable-airflow/dags.git
+        branch: test/git-sync-1
         credentials:
           sshPrivateKeySecretName: git-sync-ssh
 {% endif %}
 {% if test_scenario['values']['access'] == 'https' %}
       - repo: https://github.com/stackable-airflow/dags
+        branch: test/git-sync-0
+        credentials:
+          basicAuthSecretName: git-credentials
+      - repo: https://github.com/stackable-airflow/dags
+        branch: test/git-sync-1
         credentials:
           basicAuthSecretName: git-credentials
 {% endif %}
 {% if test_scenario['values']['executor'] == 'celery' %}
         # Just setting some values to increase the test coverage (defaults should work just fine)
-        branch: main
         wait: 5s
 {% endif %}
         gitSyncConf: