Feature: Spark retry docker pull (#672)

* retry docker pulls

* change order of pool, job, storage creation to reduce conflicts

* add error message on docker-compose curl failure

aztk/client/cluster/helpers/create.py

@@ -22,12 +22,9 @@ def create_pool_and_job_and_table(
         :param VmImageModel: the type of image to provision for the cluster
         :param wait: wait until the cluster is ready
     """
-    # update storage with the necessary values
+    # save cluster configuration in storage
     core_cluster_operations.get_cluster_data(cluster_conf.cluster_id).save_cluster_config(cluster_conf)
 
-    if cluster_conf.scheduling_target != models.SchedulingTarget.Any:
-        core_cluster_operations.create_task_table(cluster_conf.cluster_id)
-
     # reuse pool_id as job_id
     pool_id = cluster_conf.cluster_id
     job_id = cluster_conf.cluster_id
@@ -71,4 +68,8 @@ def create_pool_and_job_and_table(
     # Add job to batch
     core_cluster_operations.batch_client.job.add(job)
 
+    # create storage task table
+    if cluster_conf.scheduling_target != models.SchedulingTarget.Any:
+        core_cluster_operations.create_task_table(cluster_conf.cluster_id)
+
     return helpers.get_cluster(cluster_conf.cluster_id, core_cluster_operations.batch_client)

aztk/client/job/helpers/submit.py

@@ -29,9 +29,6 @@ def submit_job(
     core_job_operations.get_cluster_data(job_configuration.id).save_cluster_config(
         job_configuration.to_cluster_config())
 
-    if job_configuration.scheduling_target != models.SchedulingTarget.Any:
-        core_job_operations.create_task_table(job_configuration.id)
-
     # get a verified node agent sku
     sku_to_use, image_ref_to_use = helpers.select_latest_verified_vm_image_with_node_agent_sku(
         vm_image_model.publisher, vm_image_model.offer, vm_image_model.sku, core_job_operations.batch_client)
@@ -84,4 +81,7 @@ def submit_job(
 
     core_job_operations.batch_client.job_schedule.add(setup)
 
+    if job_configuration.scheduling_target != models.SchedulingTarget.Any:
+        core_job_operations.create_task_table(job_configuration.id)
+
     return core_job_operations.batch_client.job_schedule.get(job_schedule_id=job_configuration.id)

aztk/node_scripts/setup_host.sh

@@ -42,8 +42,11 @@ install_prerequisites () {
 
 install_docker_compose () {
     echo "Installing Docker-Compose"
+    url=https://github.com/docker/compose/releases/download/1.19.0/docker-compose-`uname -s`-`uname -m`
     for i in {1..5}; do
-        sudo curl -L https://github.com/docker/compose/releases/download/1.19.0/docker-compose-`uname -s`-`uname -m` -o /usr/local/bin/docker-compose && break || sleep 2; 
+        sudo curl -L $url -o /usr/local/bin/docker-compose && break ||
+        echo "ERROR: failed to download docker-compose ... retrying in $($i**2) seconds" &&
+        sleep $i**2;
     done
     sudo chmod +x /usr/local/bin/docker-compose
     echo "Finished installing Docker-Compose"
@@ -59,7 +62,12 @@ pull_docker_container () {
         docker login $DOCKER_ENDPOINT --username $DOCKER_USERNAME --password $DOCKER_PASSWORD
     fi
 
-    docker pull $docker_repo_name
+
+    for i in {1..5}; do
+        docker pull $docker_repo_name && break ||
+        echo "ERROR: docker pull $docker_repo_name failed ... retrying after $($i**2) seconds" &&
+        sleep $i**2;
+    done
     echo "Finished pulling $docker_repo_name"
 }