MySQL monitor, troubleshoot performance bottlenecks

 

 

The role of monitoring

  Business / Database Service is available

    Can external services

    In the process allows, but did not listen to the network, or does not properly authorized, or network failure

    We can not only monitor the process starts or not, whether to listen to the network

    Best simulate the business logic to monitor

    In addition to the business logic to complete availability monitoring can also monitor data reliability

  Data reliability

    Whether the service is reliable data

    Service is available, does not mean the data is correct

    Possible misuse delete data, or other unexpected causes of data loss

    Or master-slave replication latency, leading to the latest data can not be read from the server to the

    By pseudo-random business logic to verify the reliability of data

  Server & MySQL instances appear high debt

    Service is available, but the response is slow, in fact, equal unavailable

    When the response is very slow, impatient users have been refreshed, more likely to cause a storm

    When the entire system needs immediate attention long response, the number of transactions per second

  Monitoring can focus on the current availability of historical data and trends

  Availability warning

  Historical Trend

    Learn online computing resource usage  

    As computing resources expansion / contraction reference

    As a result of the optimization work records show

  Performance through real-time transaction data change the perception of changes in the business

  Whether the database server performance trends to determine adequate resources

 

Monitoring of key indicators

  Normal operation Summary

  CPU: %user, %sys, %idel, %iowait

  Memory: free, cached, swap, and if there is a memory leak OOM

  I / O: iops, throughput, latency, utilization (% util)

  NIC: Throughput (packet transceiver frequency of particular concern)

 

System monitoring

  General Tools

    top, free, ps ,df

    sysstat (sar,mpstat, iostat) \ dstat \ iotop

      sar -u cpu state

      sar -d disk state

      iostat -dkx / dmx  

      mpstat view cpu interrupt uneven mpstat -I SUM -P ALL 1 10 (if not interrupted balanced Description Hard network card default binding to a certain cpu, cpu can use all changed)

    netstat, ethstatus, arping

  Other tools

    perf

    pstack

  

 

Guess you like

Origin www.cnblogs.com/yujiaershao/p/11387787.html