Merge pull request #923 from barakda/squid

Squid branch update

.env

@@ -1,7 +1,7 @@
 # Globals
-VERSION="1.3.2"
-CEPH_VERSION="18.2.4"
-SPDK_VERSION="24.01"
+VERSION="1.3.3"
+CEPH_VERSION="19.2.0"
+SPDK_VERSION="24.09"
 CONTAINER_REGISTRY="quay.io/ceph"
 QUAY_SPDK="${CONTAINER_REGISTRY}/spdk"
 QUAY_CEPH="${CONTAINER_REGISTRY}/vstart-cluster"

.github/workflows/build-container.yml

@@ -737,7 +737,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        test: ["sanity", "ns_lb_change", "state_transitions", "state_transitions_both_gws", "state_transitions_loop", "state_transitions_rand_loop", "late_registration", "late_registration_loop", "4gws", "4gws_loop", "4gws_create_delete", "4gws_create_delete_loop", "namespaces", "namespaces_loop", "mtls", "notify", "ceph_status"]
+        test: ["sanity", "ns_lb_change", "state_transitions", "state_transitions_both_gws", "state_transitions_loop", "state_transitions_rand_loop", "late_registration", "late_registration_loop", "4gws", "4gws_loop", "4gws_create_delete", "4gws_create_delete_loop", "namespaces", "namespaces_loop", "mtls", "notify", "ceph_status",  "main_exit"]
     runs-on: ubuntu-latest
     env:
       HUGEPAGES: 1024  # 4 spdk instances

control/__main__.py

@@ -8,14 +8,10 @@
 #
 
 import argparse
-import signal
 from .server import GatewayServer
 from .config import GatewayConfig
 from .utils import GatewayLogger
 
-def sigterm_handler(signum, frame):
-    raise SystemExit(f"Gateway process terminated")
-
 if __name__ == '__main__':
     parser = argparse.ArgumentParser(prog="python3 -m control",
                                      description="Manage NVMe gateways",
@@ -32,10 +28,6 @@ def sigterm_handler(signum, frame):
     gw_logger = GatewayLogger(config)
     config.display_environment_info(gw_logger.logger)
     config.dump_config_file(gw_logger.logger)
-    try:
-        with GatewayServer(config) as gateway:
-            signal.signal(signal.SIGTERM, sigterm_handler)
-            gateway.serve()
-            gateway.keep_alive()
-    except SystemExit:
-        pass
+    with GatewayServer(config) as gateway:
+        gateway.serve()
+        gateway.keep_alive()

control/server.py

@@ -36,6 +36,12 @@
 from .cephutils import CephUtils
 from .prometheus import start_exporter
 
+def sigterm_handler(signum, frame):
+    """Handle SIGTERM, runs when a gateway is terminated gracefully."""
+    logger = GatewayLogger().logger
+    logger.info(f"GatewayServer: SIGTERM received {signum=}")
+    raise SystemExit(0)
+
 def sigchld_handler(signum, frame):
     """Handle SIGCHLD, runs when a child process, like the spdk, terminates."""
     logger = GatewayLogger().logger
@@ -179,6 +185,9 @@ def serve(self):
         # install SIGCHLD handler
         signal.signal(signal.SIGCHLD, sigchld_handler)
 
+        # install SIGTERM handler
+        signal.signal(signal.SIGTERM, sigterm_handler)
+
         # Start monitor client
         self._start_monitor_client()
 
@@ -574,7 +583,9 @@ def keep_alive(self):
                 consecutive_ping_failures += 1
                 if consecutive_ping_failures >= allowed_consecutive_spdk_ping_failures:
                     self.logger.critical(f"SPDK ping failed {consecutive_ping_failures} times, aborting")
-                    break
+                    raise SystemExit(f"SPDK ping failed, quitting gateway")
+                else:
+                    self.logger.warning(f"SPDK ping failed {consecutive_ping_failures} times, will keep trying")
             else:
                 consecutive_ping_failures = 0
 

pyproject.toml

@@ -4,7 +4,8 @@ build-backend = "pdm.backend"
 
 [project]
 name = "ceph-nvmeof"
-version = "1.3.1"
+version = "1.3.3"
+
 description = "Service to provide Ceph storage over NVMe-oF protocol"
 readme = "README.md"
 requires-python = "~=3.9"

tests/ha/main_exit.sh

@@ -0,0 +1,111 @@
+#!/bin/sh
+set -xe
+SCALE=1
+POOL="${RBD_POOL:-rbd}"
+
+background_task() {
+
+  # Give gateway some time
+  sleep 5
+
+  # Waiting for the ceph container to become healthy
+  while true; do
+    container_status=$(docker inspect --format='{{.State.Health.Status}}' ceph)
+    if [ "$container_status" = "healthy" ]; then
+      # success
+      break
+    else
+      # Wait for a specific time before checking again
+      sleep 1
+      printf .
+    fi
+  done
+  echo ✅ ceph is healthy
+
+  echo ℹ️  Running processes of services
+  docker compose top
+
+  echo ℹ️  Send nvme-gw create for all gateways
+  GW_GROUP=''
+  i=1 # a single gw index
+  GW_NAME=$(docker ps --format '{{.ID}}\t{{.Names}}' | grep -v discovery | awk '$2 ~ /nvmeof/ && $2 ~ /'$i'/ {print $1}')
+  echo  📫 nvme-gw create gateway: \'$GW_NAME\' pool: \'$POOL\', group: \'$GW_GROUP\'
+  docker compose exec -T ceph ceph nvme-gw create $GW_NAME $POOL "$GW_GROUP"
+
+  echo ℹ️  Wait for gateway to be ready
+  while true; do
+    sleep 1  # Adjust the sleep duration as needed
+    container_status=$(docker inspect -f '{{.State.Status}}' "$GW_NAME")
+    if [ "$container_status" == "running" ]; then
+      echo "Container $i $GW_NAME is now running."
+    else
+      echo "Container $i $GW_NAME is still not running. Waiting..."
+      continue
+    fi
+    GW_IP="$(docker inspect -f '{{range .NetworkSettings.Networks}}{{.IPAddress}}{{end}}' "$GW_NAME")"
+    if docker compose run --rm nvmeof-cli $CLI_TLS_ARGS --server-address $GW_IP --server-port 5500 get_subsystems 2>&1 | grep -i failed; then
+      echo "Container $i $GW_NAME $GW_IP no subsystems. Waiting..."
+      continue
+    fi
+    echo "Container $i $GW_NAME $GW_IP subsystems:"
+    docker compose run --rm nvmeof-cli $CLI_TLS_ARGS --server-address $GW_IP --server-port 5500 get_subsystems
+    break
+  done
+
+  # Signal to send (e.g., SIGTERM or SIGKILL)
+  SIGNAL="SIGABRT"
+
+  # Get the PID of monitor_client inside the container
+  PID=$(docker exec "$GW_NAME" sh -c "for pid in /proc/*; do
+        if [ -f \"\$pid/comm\" ] && grep -q 'ceph-nvmeof-mon' \"\$pid/comm\"; then
+            echo \$(basename \$pid)
+            break
+        fi
+    done")
+
+  if [ -n "$PID" ]; then
+    echo "ℹ️  Sending $SIGNAL to monitor_client (PID: $PID) in $GW_NAME..."
+    docker exec "$GW_NAME" kill -s "$SIGNAL" "$PID"
+  else
+    echo "❌ monitor_client process not found in $GW_NAME."
+    exit 1
+  fi
+
+}
+
+##
+## MAIN
+##
+
+background_task &
+TASK_PID=$!  # Capture the PID of the background task
+
+echo ℹ️  Starting $SCALE nvmeof gateways
+docker compose up --remove-orphans --scale nvmeof=$SCALE nvmeof
+GW_NAME=$(docker ps -a --format '{{.ID}}\t{{.Names}}' | grep -v discovery | awk '$2 ~ /nvmeof/ && $2 ~ /'$i'/ {print $1}')
+docker inspect "$GW_NAME"
+exit_code=$(docker inspect --format='{{.State.ExitCode}}' "$GW_NAME")
+
+# expect exit code 1
+if [ $exit_code -eq 1 ]; then
+    echo ✅  gateway returned exit code 1, exiting with success.
+else
+    echo ❌  gateway returned exit code $exit_code, exiting with failure.
+    exit 1  # Failure exit code
+fi
+
+# Wait for the background task to finish
+wait $TASK_PID  # Wait for the specific PID to complete
+background_task_exit_code=$?    # Capture the exit code of the background task
+
+# Check the exit code and print the result
+if [ $background_task_exit_code -eq 0 ]; then
+    echo ✅ background task completed successfully
+else
+    echo ❌ background task failed with exit code: $background_task_exit_code
+fi
+
+# Exit with the same code as the background task
+exit $background_task_exit_code
+
+

tests/ha/setup_main_exit.sh

@@ -0,0 +1,2 @@
+set -e
+echo ℹ️  Skipping setup for this test

tests/ha/start_up_main_exit.sh

@@ -0,0 +1,2 @@
+set -e
+echo ℹ️  Skipping start up for this test

tests/ha/wait_gateways.sh

@@ -1,3 +1,5 @@
+#!/bin/sh
+set -ex
 SCALE=2
 echo CLI_TLS_ARGS $CLI_TLS_ARGS
 # Check if argument is provided
@@ -16,7 +18,7 @@ for i in $(seq $SCALE); do
     sleep 1  # Adjust the sleep duration as needed
     GW_NAME=$(docker ps --format '{{.ID}}\t{{.Names}}' | awk '$2 ~ /nvmeof/ && $2 ~ /'$i'/ {print $1}')
     container_status=$(docker inspect -f '{{.State.Status}}' "$GW_NAME")
-    if [ "$container_status" == "running" ]; then
+    if [ "$container_status" = "running" ]; then
       echo "Container $i $GW_NAME is now running."
     else
       echo "Container $i $GW_NAME is still not running. Waiting..."

tests/ha/wait_gateways_main_exit.sh

@@ -0,0 +1,2 @@
+set -e
+echo ℹ️  Skipping wait gateways up for this test