ah

Signed-off-by: Judy Ng <[email protected]>

tests/integration-tests/configs/scaling_stress_test.yaml

@@ -2,13 +2,7 @@ test-suites:
   performance_tests:
     test_scaling.py::test_scaling_stress_test:
       dimensions:
-        - regions: [ "us-east-1" ]
-          instances: [ "c5.large" ]
-          oss: [ "alinux2" ]
-          schedulers: [ "slurm" ]
-    test_scaling.py::test_static_scaling_stress_test:
-      dimensions:
-        - regions: [ "us-east-1" ]
-          instances: [ "c5.large" ]
+        - regions: [ "euw1-az2" ]
+          instances: [ "p3.2xlarge" ]
           oss: [ "alinux2" ]
           schedulers: [ "slurm" ]

tests/integration-tests/tests/common/scaling/get_bootstrap_errors.sh

@@ -19,7 +19,7 @@ touch "bootstrap_errors.txt"
 # Find a log message like:
 # ... WARNING - Node bootstrap error: Node queue-0-dy-compute-resource-0-1690(192.168.90.197) ...
 # and get the IP address
-sudo cat ${CLUSTERMGTD_LOG} | grep -i "Node bootstrap error" | awk -F"[()]" '{print $2}' | while read -r ip_address ; do
+sudo cat ${CLUSTERMGTD_LOG} | grep -i "no corresponding instance in EC2 for node" | awk -F"[()]" '{print $2}' | while read -r ip_address ; do
   if ! grep -q "${ip_address}" "bootstrap_errors.txt"; then
     echo "${ip_address}" >> "bootstrap_errors.txt"
   fi

tests/integration-tests/tests/common/scaling_common.py

@@ -94,13 +94,13 @@ def get_bootstrap_errors(remote_command_executor: RemoteCommandExecutor, cluster
         # Since the same cluster is re-used for multiple scale up tests, the script may find the same bootstrap error
         # multiple times and then get the wrong instance logs since the IP address would be attached to a new instance.
         # Therefore, only write the compute node logs for the IP address if the file doesn't exist yet.
-        if _check_no_node_log_exists_for_ip_address("bootstrap_errors", ip_address):
+        if _check_no_node_log_exists_for_ip_address(path, ip_address):
             try:
                 logging.warning(f"Compute node with IP address {ip_address} had bootstrap errors. Getting instance id...")
                 # Get the latest launched instance with the IP address since the most recent one should have the error
                 instance_id = _sort_instances_by_launch_time(client.describe_instances(
                     Filters=[{"Name": "private-ip-address", "Values": [ip_address]}]
-                ))[-1]
+                ))[-1]["InstanceId"]
                 logging.warning(f"Instance {instance_id} had bootstrap errors. Check the logs for details.")
                 compute_node_log = client.get_console_output(InstanceId=instance_id)["Output"]
                 with open(os.path.join(path, f"{ip_address}-{cluster_name}-{instance_id}-{region}-log.txt", "w")) as f:

tests/integration-tests/tests/performance_tests/test_scaling.py

@@ -82,7 +82,7 @@ def _get_scaling_time(capacity_time_series: list, timestamps: list, scaling_targ
 
 
 @pytest.mark.usefixtures("scheduler")
-@pytest.mark.parametrize("scaling_strategy", ["all-or-nothing", "best-effort"])
+@pytest.mark.parametrize("scaling_strategy", ["all-or-nothing"])
 def test_scaling_stress_test(
     test_datadir,
     instance,