Summary of server drops caused by network card traffic

Recently received a line of a DB server ping drops, packet loss rate has been around 30%. Through the Zabbix monitoring to view the server CPU, memory is normal, network card traffic is not high, basically around 100M.
650) this.width=650; "Src=" Http://s3.51cto.com/wyfs02/M00/85/75/wKiom1ekGFCBE0nWAAHfgGoPDTk656.jpg-wh_500x0-wm_3 -wmp_4-s_2484793242.jpg "title=" icmpping.jpg "alt=" Wkiom1ekgfcbe0nwaahfggopdtk656.jpg-wh_50 "/>

First confirm that the server hardware is normal, because no hardware alarm received. The login server confirms that the hardware information is normal (hard disk, cache card, memory, etc.) through the HP management tool.
The second step in troubleshooting system problems, through the Top,ps and other commands have not found anything unusual, basically eliminate system problems.
The third step to check the server uplink monitoring machine port traffic, is also very normal, due to receive only this one server alarm, but also ruled out the switch fault problem.
Finally to colleagues consulted the server hosting business type, every 2 minutes will synchronize a large number of data files to the server, and then use the SAR command to view the network card traffic, found that the sending traffic instantaneous in 120,000 kb/s, converted to B/s basically in 940-950mb/s, means that gigabit network card traffic is basically full , the server ping drops are not caused.
650) this.width=650; "Src=" Http://s3.51cto.com/wyfs02/M00/85/75/wKioL1ekGHTRDUfnAAGPak_j6LQ842.jpg-wh_500x0-wm_3 -wmp_4-s_3458675798.jpg "title=" sar.jpg "alt=" Wkiol1ekghtrdufnaagpak_j6lq842.jpg-wh_50 "/>
Since my monitoring is every 5 minutes to catch, so the corresponding server instantaneous high traffic is not acquired, but also to optimize the monitoring interval.

Here is a summary of the options commonly used by the SAR command, the common format of the SAR command line is as follows:

SAR option sample time interval output number

1) View the CPU information, 1 means 1 seconds to take the value, 2 means to collect 2 times of data.

[[email protected] ~]# sar -u 1 2linux 2.6.32-358.el6.x86_64  (monitor)      08/05/16     _x86_64_     (24  CPU)  10:51:39        CPU     %user      %nice   %system   %iowait    %steal      %idle10:51:40        all       0.08      0.00      0.17       0.00      0.00      99.7510:51:41        all      0.21       0.00      0.21       0.00      0.00     99.58average:         all      0.15      0.00       0.19      0.00      0.00      99.67

Output Item Description:

The CPU all represents the average of the statistics for all CPUs. %user shows the percentage of total CPU time that is running at the user level (application). %nice displays the percentage of total CPU time that is consumed at the user level for the nice operation. %system the percentage of total CPU time that is used to run at the core level (kernel). %iowait displays the percentage of the total CPU time that waits for I/O operations to occupy. The percentage of the%steal hypervisor (hypervisor) that waits for a virtual CPU to serve another virtual process. %idle shows the percentage of CPU idle time that takes up the total CPU time.

2) View the network interface information.

[[email protected] ~]# sar -n dev 1 2linux 2.6.32-358.el6.x86_64  (monitor)      08/05/16     _x86_64_     ( 24 CPU)  11:04:22        IFACE   rxpck/s    txpck/s    rxkB/s    txkB/s   rxcmp/s    txcmp/s  rxmcst/s11:04:23            lo      0.00      0.00       0.00      0.00      0.00       0.00      0.0011:04:23          eth0    140.40    170.71      98.07     84.00      0.00       0.00      2.0211:04:23         eth1       0.00      0.00       0.00      0.00      0.00       0.00      0.0011:04:23          eth2      0.00      0.00       0.00      0.00       0.00      0.00      0.0011:04:23          eth3      0.00       0.00       0.00      0.00      0.00       0.00      0.00 11:04:23         IFACE   rxpck/s   txpck/s     rxkb/s    txkb/s   rxcmp/s   txcmp/s  rxmcst /s11:04:24           lo       0.00      0.00      0.00       0.00      0.00      0.00       0.0011:04:24         eth0      40.59     26.73     41.62       4.17      0.00      0.00       0.9911:04:24         eth1       0.00      0.00      0.00       0.00      0.00      0.00       0.0011:04:24         eth2       0.00      0.00       0.00      0.00      0.00       0.00      0.0011:04:24          eth3      0.00      0.00       0.00      0.00      0.00       0.00      0.00 Average:         IFACE   rxpck/s   txpck/s    rxkB/s     txkB/s   rxcmp/s   txcmp/s  rxmcst/sAverage:            lo      0.00       0.00      0.00      0.00       0.00      0.00       0.00Average:         eth0      90.00     98.00     69.56     43.69       0.00      0.00      1.50Average:          eth1      0.00       0.00      0.00      0.00       0.00      0.00       0.00average:         eth2      0.00       0.00      0.00       0.00      0.00      0.00       0.00Average:         eth3       0.00      0.00      0.00       0.00      0.00      0.00       0.00

Output Item Description:

IFACE Network device name rxpck/s The total number of packets received per second TXPCK/S the total number of packets transmitted per second rxkb/s bytes received per second (byte) total txkb/s bytes transferred per second (byte) rxcmp/ s total number of received compressed packets per second TXCMP/S total number of transmit compressed packets per second RXMCST/S total number of multicast (multicast) packets received per second 3) view disk 1/0 information.

[[email protected] ~]# sar -b 1 2linux 2.6.32-358.el6.x86_64  (monitor)      08/05/16     _x86_64_     (24  CPU)  11:07:55          tps       rtps      wtps   bread/s   bwrtn/s11:0 7:56        11.11      0.00      11.11      0.00    129.2911:07:57          6.93      0.00       6.93      0.00     63.37average:          9.00      0.00       9.00      0.00     96.00 

Output Item Description:

TPS Total I/O transfer per second of physical device Rtps total amount of data read from physical device per second Wtps total amount of data written to physical device per second bread/s the amount of data read from the physical device per second, in blocks/sbwrtn/s per second The amount of data written by the device, in blocks per/s

Summary: In the system operation and maintenance process, the general focus on the following indicators of the server.
CPU Usage: If the server CPU usage exceeds 80-85%, the server CPU processing power is busy and needs to improve CPU performance.
CPU iowait: If the value of the server CPU iowait is greater than 5-10%, the disk I/O has a bottleneck and needs to increase the read and write speed of the hard disk.
Network card traffic: Network card traffic and the uplink switch and the server network card are related. If the system and network are normal, the server drops packets, you should consider whether the network card throughput is up to the upper limit of packet loss.

This article is from the "Simple Dreamer" blog, please be sure to keep this source http://sfzhang88.blog.51cto.com/4995876/1834696

Summary of server drops caused by network card traffic