push disk io for client-linux.py

clients/client-linux.py

@@ -214,9 +214,63 @@ def _net_speed():
         time.sleep(INTERVAL)
 
 def _disk_io():
-    # todo, 这里希望能有人贡献下代码
-    diskIO["read"] = 0
-    diskIO["write"] = 0
+    '''
+    磁盘IO：因为IOPS原因，SSD和HDD、包括RAID卡，ZFS等阵列技术。IO对性能的影响还需要结合自身服务器情况来判断。
+    比如我这里是机械硬盘，大量做随机小文件读写，那么很低的读写也就能造成硬盘长时间的等待。
+    如果这里做连续性IO，那么普通机械硬盘写入到100Mb/s，那么也能造成硬盘长时间的等待。
+    磁盘读写有误差：4k，8k ，https://stackoverflow.com/questions/34413926/psutil-vs-dd-monitoring-disk-i-o
+    :return:
+    '''
+    # pre pid snapshot
+    snapshot_first = {}
+    # next pid snapshot
+    snapshot_second = {}
+    # read count snapshot
+    snapshot_read = 0
+    # write count snapshot
+    snapshot_write = 0
+    # process snapshot
+    pid_snapshot = [str(i) for i in os.listdir("/proc") if i.isdigit() is True]
+    for pid in pid_snapshot:
+        try:
+            with open("/proc/{}/io".format(pid)) as f:
+                pid_io = {}
+                for line in f.readlines():
+                    if "read_bytes" in line:
+                        pid_io["read"] = int(line.split("read_bytes:")[-1].strip())
+                    elif "write_bytes" in line and "cancelled_write_bytes" not in line:
+                        pid_io["write"] = int(line.split("write_bytes:")[-1].strip())
+                pid_io["name"] = open("/proc/{}/comm".format(pid), "r").read().strip()
+                snapshot_first[pid] = pid_io
+        except:
+            if pid in snapshot_first:
+                snapshot_first.pop(pid)
+
+    time.sleep(INTERVAL)
+
+    for pid in pid_snapshot:
+        try:
+            with open("/proc/{}/io".format(pid)) as f:
+                pid_io = {}
+                for line in f.readlines():
+                    if "read_bytes" in line:
+                        pid_io["read"] = int(line.split("read_bytes:")[-1].strip())
+                    elif "write_bytes" in line and "cancelled_write_bytes" not in line:
+                        pid_io["write"] = int(line.split("write_bytes:")[-1].strip())
+                pid_io["name"] = open("/proc/{}/comm".format(pid), "r").read().strip()
+                snapshot_second[pid] = pid_io
+        except:
+            if pid in snapshot_first:
+                snapshot_first.pop(pid)
+            if pid in snapshot_second:
+                snapshot_second.pop(pid)
+
+    for k, v in snapshot_first.items():
+        if snapshot_first[k]["name"] == snapshot_second[k]["name"] and snapshot_first[k]["name"] != "bash":
+            snapshot_read += (snapshot_second[k]["read"] - snapshot_first[k]["read"])
+            snapshot_write += (snapshot_second[k]["write"] - snapshot_first[k]["write"])
+    diskIO["read"] = snapshot_read
+    diskIO["write"] = snapshot_write
 
 def get_realtime_data():
     '''

clients/client-psutil.py

@@ -201,6 +201,10 @@ def _disk_io():
         after the interval.
         Return a tuple including all currently running processes
         sorted by IO activity and total disks I/O activity.
+    磁盘IO：因为IOPS原因，SSD和HDD、包括RAID卡，ZFS等。IO对性能的影响还需要结合自身服务器情况来判断。
+    比如我这里是机械硬盘，大量做随机小文件读写，那么很低的读写也就能造成硬盘长时间的等待。
+    如果这里做连续性IO，那么普通机械硬盘写入到100Mb/s，那么也能造成硬盘长时间的等待。
+    磁盘读写有误差：4k，8k ，https://stackoverflow.com/questions/34413926/psutil-vs-dd-monitoring-disk-i-o
     """
     while True:
         # first get a list of all processes and disk io counters