Fix for redis memory check failure after link flap and also sometimes…

… cpu usage high failure (#15732)

Description of PR
Redis memory check result is not a stable value using "redis-cli info memory | grep used_memory_human". It's found on a stable system (BGO converged, no port flapping etc), the above check could have memory usage difference by more than 0.2M.

Followings are CLI output from 202405 and 202205.

202405:
admin@ixre-egl-board30: redis-cli info memory | grep used_memory_human | sed -e 's/.:(.)M/\1/'
2.64
admin@ixre-egl-board30: redis-cli info memory | grep used_memory_human | sed -e 's/.:(.)M/\1/'
2.74
admin@ixre-egl-board30: redis-cli info memory | grep used_memory_human | sed -e 's/.:(.)M/\1/'
2.52

202205:
admin@ixre-egl-board64: redis-cli info memory | grep used_memory_human | sed -e 's/.:(.)M/\1/'
6.02
admin@ixre-egl-board64: redis-cli info memory | grep used_memory_human | sed -e 's/.:(.)M/\1/'
6.26
admin@ixre-egl-board64: redis-cli info memory | grep used_memory_human | sed -e 's/.:(.)M/\1/'
6.14

We can see that 202405 has some memory optimization for redis and it's not using as much memory as 202205. 0.2M memory usage difference could easily reach the memory usage threshold of 5% in 202405.

Solution is to get the average redis memory usage before and after link flap. using 5 seconds interval and 5 times query and then get the average memory usage for redis. Also make the threshold to 10% from 5%. With this fix it's found that the redis memory check will not fail for 2405 after link flap.

This commit also provide a fix for sometimes CPU utilization check failed for orchagent after link flap. The reason is in scaling setup (34k routes) orchagent takes more time to calm down.

Summary:
Fixes # (issue) #15733

Approach
What is the motivation for this PR?
Fix test failures

How did you verify/test it?
OC tests run with the fix. Did not see the test failed.

co-authorized by: jianquanye@microsoft.com

tests/platform_tests/link_flap/link_flap_utils.py

@@ -4,6 +4,7 @@
 import time
 import logging
 import random
+import time
 
 from tests.common.platform.device_utils import fanout_switch_port_lookup
 from tests.common.utilities import wait_until
@@ -236,3 +237,40 @@ def check_bgp_routes(dut, start_time_ipv4_route_counts, start_time_ipv6_route_co
     incr_ipv4_route_counts = abs(int(float(start_time_ipv4_route_counts)) - int(float(routesv4)))
     incr_ipv6_route_counts = abs(int(float(start_time_ipv6_route_counts)) - int(float(routesv6)))
     return incr_ipv4_route_counts < MAX_DIFF and incr_ipv6_route_counts < MAX_DIFF
+
+
+def get_avg_redis_mem_usage(duthost, interval, num_times):
+    """
+        Redis memory usage is not a stable value. It's fluctuating even when the device is stable stage.
+        202205 has larger redis memory usage (~ 5.5M) so the fluctuation of 0.2M is not an issue.
+        With 202405 redis memory usage is optimized (~ 2.5M) and 0.2M usage could make the test fail
+        if memory threshold is 5%.
+
+        This API returns the average radis memory usage during a period.
+        Args:
+            duthost: DUT host object
+            interval: time interval to wait for next query
+            num_times: number of times to query
+        """
+    logger.info("Checking average redis memory usage")
+    cmd = r"redis-cli info memory | grep used_memory_human | sed -e 's/.*:\(.*\)M/\1/'"
+    redis_memory = 0.0
+    for i in range(num_times):
+        redis_memory += float(duthost.shell(cmd)["stdout"])
+        time.sleep(interval)
+    return float(redis_memory/num_times)
+
+
+def validate_redis_memory_increase(tbinfo, start_mem, end_mem):
+    # Calculate diff in Redis memory
+    incr_redis_memory = end_mem - start_mem
+    logging.info("Redis memory usage difference: %f", incr_redis_memory)
+
+    # Check redis memory only if it is increased else default to pass
+    if incr_redis_memory > 0.0:
+        percent_incr_redis_memory = (incr_redis_memory / start_mem) * 100
+        logging.info("Redis Memory percentage Increase: %d", percent_incr_redis_memory)
+        incr_redis_memory_threshold = 15 if tbinfo["topo"]["type"] in ["m0", "mx"] else 10
+        if percent_incr_redis_memory >= incr_redis_memory_threshold:
+            return False
+    return True

tests/platform_tests/link_flap/test_cont_link_flap.py

@@ -16,7 +16,7 @@
 from tests.common.helpers.assertions import pytest_assert, pytest_require
 from tests.common import port_toggle
 from tests.platform_tests.link_flap.link_flap_utils import build_test_candidates, toggle_one_link,\
-    check_orch_cpu_utilization, check_bgp_routes
+    check_orch_cpu_utilization, check_bgp_routes, , get_avg_redis_mem_usage, validate_redis_memory_increase
 from tests.common.utilities import wait_until
 from tests.common.devices.eos import EosHost
 from tests.common.devices.sonic import SonicHost
@@ -64,7 +64,7 @@ def test_cont_link_flap(self, request, duthosts, nbrhosts, enum_rand_one_per_hws
             3.) Watch for memory (show system-memory), FRR daemons memory(vtysh -c "show memory bgp/zebra"),
                 orchagent CPU Utilization and Redis_memory.
 
-        Pass Criteria: All routes must be re-learned with < 5% increase in Redis/FRR memory usage and
+        Pass Criteria: All routes must be re-learned with < 10% increase in Redis/FRR memory usage and
             ORCH agent CPU consumption below threshold after 3 mins after stopping flaps.
         """
         duthost = duthosts[enum_rand_one_per_hwsku_frontend_hostname]
@@ -76,9 +76,8 @@ def test_cont_link_flap(self, request, duthosts, nbrhosts, enum_rand_one_per_hws
         logging.info("Memory Status at start: %s", memory_output)
 
         # Record Redis Memory at start
-        start_time_redis_memory = duthost.shell(
-            r"redis-cli info memory | grep used_memory_human | sed -e 's/.*:\(.*\)M/\1/'")["stdout"]
-        logging.info("Redis Memory: %s M", start_time_redis_memory)
+        start_time_redis_memory = get_avg_redis_mem_usage(duthost, 5, 5)
+        logging.info("Redis Memory: %f M", start_time_redis_memory)
 
         # Record ipv4 route counts at start
         sumv4, sumv6 = duthost.get_ip_route_summary(skip_kernel_tunnel=True)
@@ -207,26 +206,17 @@ def test_cont_link_flap(self, request, duthosts, nbrhosts, enum_rand_one_per_hws
             logging.info("Orchagent PID {0} CPU Util at end: {1}".format(pid, util))
 
         # Record Redis Memory at end
-        end_time_redis_memory = duthost.shell(
-            r"redis-cli info memory | grep used_memory_human | sed -e 's/.*:\(.*\)M/\1/'")["stdout"]
-        logging.info("Redis Memory at start: %s M", start_time_redis_memory)
-        logging.info("Redis Memory at end: %s M", end_time_redis_memory)
-
-        # Calculate diff in Redis memory
-        incr_redis_memory = float(end_time_redis_memory) - float(start_time_redis_memory)
-        logging.info("Redis absolute  difference: %d", incr_redis_memory)
-
-        # Check redis memory only if it is increased else default to pass
-        if incr_redis_memory > 0.0:
-            percent_incr_redis_memory = (incr_redis_memory / float(start_time_redis_memory)) * 100
-            logging.info("Redis Memory percentage Increase: %d", percent_incr_redis_memory)
-            incr_redis_memory_threshold = 10 if tbinfo["topo"]["type"] in ["m0", "mx"] else 5
-            pytest_assert(percent_incr_redis_memory < incr_redis_memory_threshold,
-                          "Redis Memory Increase more than expected: {}".format(percent_incr_redis_memory))
+        end_time_redis_memory = get_avg_redis_mem_usage(duthost, 5, 5)
+        logging.info("Redis Memory at start: %f M", start_time_redis_memory)
+        logging.info("Redis Memory at end: %f M", end_time_redis_memory)
+
+        result = validate_redis_memory_increase(tbinfo, start_time_redis_memory, end_time_redis_memory)
+        pytest_assert(result, "Redis Memory Increases more than expected: start {}, end {}"
+                      .format(start_time_redis_memory, end_time_redis_memory))
 
         # Orchagent CPU should consume < orch_cpu_threshold at last.
         logging.info("watch orchagent CPU utilization when it goes below %d", orch_cpu_threshold)
-        pytest_assert(wait_until(45, 2, 0, check_orch_cpu_utilization, duthost, orch_cpu_threshold),
+        pytest_assert(wait_until(120, 5, 0, check_orch_cpu_utilization, duthost, orch_cpu_threshold),
                       "Orch CPU utilization {} > orch cpu threshold {} after link flap"
                       .format(duthost.shell("show processes cpu | grep orchagent | awk '{print $9}'")["stdout"],
                               orch_cpu_threshold))

tests/platform_tests/link_flap/test_link_flap.py

@@ -4,7 +4,8 @@
 import logging
 import pytest
 
-from tests.platform_tests.link_flap.link_flap_utils import toggle_one_link, check_orch_cpu_utilization
+from tests.platform_tests.link_flap.link_flap_utils import toggle_one_link, check_orch_cpu_utilization, \
+    get_avg_redis_mem_usage, validate_redis_memory_increase
 from tests.common.platform.device_utils import fanout_switch_port_lookup
 from tests.common.helpers.assertions import pytest_assert
 from tests.common.utilities import wait_until
@@ -36,9 +37,8 @@ def test_link_flap(request, duthosts, rand_one_dut_hostname, tbinfo, fanouthosts
     logger.info("Memory Status at start: %s", memory_output)
 
     # Record Redis Memory at start
-    start_time_redis_memory = duthost.shell(
-        r"redis-cli info memory | grep used_memory_human | sed -e 's/.*:\(.*\)M/\1/'")["stdout"]
-    logger.info("Redis Memory: %s M", start_time_redis_memory)
+    start_time_redis_memory = get_avg_redis_mem_usage(duthost, 5, 5)
+    logging.info("Redis Memory: %f M", start_time_redis_memory)
 
     # Make Sure Orch CPU < orch_cpu_threshold before starting test.
     logger.info("Make Sure orchagent CPU utilization is less that %d before link flap", orch_cpu_threshold)
@@ -75,26 +75,17 @@ def test_link_flap(request, duthosts, rand_one_dut_hostname, tbinfo, fanouthosts
     logger.info("Orchagent CPU Util at end: %s", orch_cpu)
 
     # Record Redis Memory at end
-    end_time_redis_memory = duthost.shell(
-        r"redis-cli info memory | grep used_memory_human | sed -e 's/.*:\(.*\)M/\1/'")["stdout"]
-    logger.info("Redis Memory at start: %s M", start_time_redis_memory)
-    logger.info("Redis Memory at end: %s M", end_time_redis_memory)
-
-    # Calculate diff in Redis memory
-    incr_redis_memory = float(end_time_redis_memory) - float(start_time_redis_memory)
-    logger.info("Redis absolute  difference: %d", incr_redis_memory)
-
-    # Check redis memory only if it is increased else default to pass
-    if incr_redis_memory > 0.0:
-        percent_incr_redis_memory = (incr_redis_memory / float(start_time_redis_memory)) * 100
-        logger.info("Redis Memory percentage Increase: %d", percent_incr_redis_memory)
-        incr_redis_memory_threshold = 10 if tbinfo["topo"]["type"] in ["m0", "mx"] else 5
-        pytest_assert(percent_incr_redis_memory < incr_redis_memory_threshold,
-                      "Redis Memory Increase more than expected: {}".format(percent_incr_redis_memory))
+    end_time_redis_memory = get_avg_redis_mem_usage(duthost, 5, 5)
+    logging.info("Redis Memory at start: %f M", start_time_redis_memory)
+    logging.info("Redis Memory at end: %f M", end_time_redis_memory)
+
+    result = validate_redis_memory_increase(tbinfo, start_time_redis_memory, end_time_redis_memory)
+    pytest_assert(result, "Redis Memory increases more than expected: start {}, end {}"
+                  .format(start_time_redis_memory, end_time_redis_memory))
 
     # Orchagent CPU should consume < orch_cpu_threshold at last.
     logger.info("watch orchagent CPU utilization when it goes below %d", orch_cpu_threshold)
-    pytest_assert(wait_until(45, 2, 0, check_orch_cpu_utilization, duthost, orch_cpu_threshold),
+    pytest_assert(wait_until(120, 5, 0, check_orch_cpu_utilization, duthost, orch_cpu_threshold),
                   "Orch CPU utilization {} > orch cpu threshold {} before link flap"
                   .format(duthost.shell("show processes cpu | grep orchagent | awk '{print $9}'")["stdout"],
                           orch_cpu_threshold))