Correct real time CPU Utilization calculation (sonic-net#173)

**What I did**
Correct real time CPU Utilization calculation

**Why I did it**
Now in procdockerstatsd script we iterate all processes every 2 minutes and ingest into state_db, 
https://psutil.readthedocs.io/en/latest/#psutil.cpu_percent, psutil supports both blocking and non-blocking way, and we prefer to use non-blocking way, thus now psutil.cpu_percent() used is to calculate CPU utilization since last time being triggered, current implementation will create new process object in below lines every time:
for process in psutil.process_iter(['pid', 'ppid', 'memory_percent', 'cpu_percent', 'create_time', 'cmdline']):

In order to get correct CPU utilization from last 2 minutes, we need store process object into some variable, so that every iteration will first lookup all_process_obj member and use existing process object, otherwise create new object for further use. And clean up the dead process object in case some process crashes.

**How I verified it**
Verified by UT and will check by ksoftirqd reproduced case.

scripts/procdockerstatsd

@@ -24,6 +24,7 @@ REDIS_HOSTIP = "127.0.0.1"
 
 
 class ProcDockerStats(daemon_base.DaemonBase):
+    all_process_obj = {}
 
     def __init__(self, log_identifier):
         super(ProcDockerStats, self).__init__(log_identifier)
@@ -137,10 +138,25 @@ class ProcDockerStats(daemon_base.DaemonBase):
 
     def update_processstats_command(self):
         processdata = []
-        for process in psutil.process_iter(['pid', 'ppid', 'memory_percent', 'cpu_percent', 'create_time', 'cmdline']):
+        pid_set = set()
+
+        processes_all = []
+        for process_obj in psutil.process_iter(['pid', 'ppid', 'memory_percent', 'cpu_percent', 'create_time', 'cmdline']):
+            processes_all.append(process_obj)
+        sorted_processes = sorted(processes_all, key=lambda x: x.cpu_percent(), reverse=True)
+        top_processes = sorted_processes[:1024]
+
+        for process_obj in top_processes:
             try:
+                pid = process_obj.pid
+                pid_set.add(pid)
+                # store process object and reuse for CPU utilization
+                if pid in self.all_process_obj:
+                    process = self.all_process_obj[pid]
+                else:
+                    self.all_process_obj[pid] = process_obj
+                    process = process_obj
                 uid = process.uids()[0]
-                pid = process.pid
                 ppid = process.ppid()
                 mem = process.memory_percent()
                 cpu = process.cpu_percent()
@@ -156,6 +172,14 @@ class ProcDockerStats(daemon_base.DaemonBase):
             except (psutil.NoSuchProcess, psutil.AccessDenied, psutil.ZombieProcess):
                 pass
 
+        # erase dead process
+        remove_keys = []
+        for id in self.all_process_obj:
+            if id not in pid_set:
+                remove_keys.append(id)
+        for id in remove_keys:
+            del self.all_process_obj[id]
+
         # wipe out all data before updating with new values
         self.state_db.delete_all_by_pattern('STATE_DB', 'PROCESS_STATS|*')
 

tests/procdockerstatsd_test.py

@@ -115,13 +115,21 @@ def test_update_processstats_command(self):
         valid_create_time2 = int((current_time - timedelta(days=2)).timestamp())
         # Create a list of mocked processes
         mocked_processes = [
-            MockProcess(uids=[1000], pid=1234, ppid=5678, memory_percent=10.5, cpu_percent=20.5, create_time=valid_create_time1, cmdline=['python', 'script.py'], user_time=1.5, system_time=2.0),
-            MockProcess(uids=[1000], pid=5678, ppid=0, memory_percent=5.5, cpu_percent=15.5, create_time=valid_create_time2, cmdline=['bash', 'script.sh'], user_time=3.5, system_time=4.0)
+            MockProcess(uids=[1000], pid=1234, ppid=0, memory_percent=10.5, cpu_percent=99.0, create_time=valid_create_time1, cmdline=['python', 'script.py'], user_time=1.5, system_time=2.0),
+            MockProcess(uids=[1000], pid=5678, ppid=0, memory_percent=5.5, cpu_percent=15.5, create_time=valid_create_time2, cmdline=['bash', 'script.sh'], user_time=3.5, system_time=4.0),
+            MockProcess(uids=[1000], pid=3333, ppid=0, memory_percent=5.5, cpu_percent=15.5, create_time=valid_create_time2, cmdline=['bash', 'script.sh'], user_time=3.5, system_time=4.0)
+        ]
+        mocked_processes2 = [
+            MockProcess(uids=[1000], pid=1234, ppid=0, memory_percent=10.5, cpu_percent=20.5, create_time=valid_create_time1, cmdline=['python', 'script.py'], user_time=1.5, system_time=2.0),
+            MockProcess(uids=[1000], pid=6666, ppid=0, memory_percent=5.5, cpu_percent=15.5, create_time=valid_create_time2, cmdline=['bash', 'script.sh'], user_time=3.5, system_time=4.0)
         ]
 
         with patch("procdockerstatsd.psutil.process_iter", return_value=mocked_processes) as mock_process_iter:
+            pdstatsd.all_process_obj = {1234: mocked_processes2[0],
+                                        6666: mocked_processes2[1]}
             pdstatsd.update_processstats_command()
             mock_process_iter.assert_called_once()
+        assert(len(pdstatsd.all_process_obj)== 3)
 
     @patch('procdockerstatsd.getstatusoutput_noshell_pipe', return_value=([0, 0], ''))
     def test_update_fipsstats_command(self, mock_cmd):