Troubleshooting methods for user state memory leaks

Not involving

Under normal circumstances, the system memory usage should be below 80%, otherwise it is abnormal. If the system memory utilization is stable and high, but there is no continuous increase, you need to confirm which process has high memory utilization and whether it is normal. If a pre-alarm, first-level, second-level, or third-level memory threshold alarm is generated on the device, and the remaining free memory continues to decrease and cannot be recovered, you need to confirm whether there is a memory leak, and you need to confirm whether there is a user state memory leak or the kernel State memory leak problem.

Commands used:

display memory

monitor process

display process memory

display process memory heap job jobid

display process memory heap job jobid verbose

（1）The display memory or monitor process commands are used to query the system memory statistics many times. Both commands support specific slot query:

display memroy, observe that if there is no change in the used/free data, you can confirm that there is no leak.

Monitor process and observe that if there is no change in the available data, you can confirm that there is no leakage.

If there are changes, continue to analyze.

（2）Use the display process memory or monitor process (to exit the display by q or Ctrl+C) command to view the occupancy of each process in the user mode:

Display process memory, text, and data are listed as the virtual memory space occupied, stack and dynamic are the actual physical memory occupied. Generally, when viewing memory leaks, we need to pay attention to the last dynamic column, and pay attention to whether the heap memory usage continues to increase.

Monitor process, supports sorting and viewing according to memory usage (interactive command'm'), which can easily observe the growth of memory. It should be noted that the memory here is the size of the virtual memory space of the process.

Points to note when confirming memory leaks:

1) The normal use value is required as a reference baseline.

2) Continuous testing is required to confirm that the memory does continue to grow and cannot be recovered.

Through the above steps, if it can be confirmed that the memory usage of a certain process continues to increase (the JID corresponding to the process is confirmed), we need to further check the cause of the memory leak.

Use the display process memory heap job jobid to view the heap memory consumption profile of the specified process to confirm whether there is too much allocation, and the physical heap space allocation of the process physical memory has exceeded the normal value, and then use the display process memory heap job jobid verbose to view the heap memory For detailed consumption information, you can analyze which type consumes too much memory.