Linux下如何获取等待CPU的进程数及单独统计等待进程数？

Hey there, let's break down your two Linux process metric questions clearly—these are key for debugging CPU bottlenecks!

There are a couple of practical ways to calculate this:

• Method 1: Combine vmstat and CPU core count
  First, use vmstat to get the total number of runnable processes (both running on CPU and waiting for CPU time):

  vmstat 1 1 | awk 'NR==3 {print $1}'
  

  This pulls the value from the r column (the third row is the actual data after headers). Then get your system's online CPU core count:

  nproc
  

  Since each CPU core can only run one process at a time, the number of processes waiting for CPU is simply the total runnable count minus the core count (if the total is larger than the core count; otherwise, it's 0). Here's a quick one-liner to compute it directly:

  r=$(vmstat 1 1 | awk 'NR==3 {print $1}'); cores=$(nproc); wait=$(( r > cores ? r - cores : 0 )); echo "Processes waiting for CPU: $wait"
  

• Method 2: Verify with ps
  You can count all processes in the runnable state (R state) using ps:

  ps -eo state | grep -c '^R'
  

  This number matches the r value from vmstat. Subtract your CPU core count from this number, and you'll get the count of processes waiting for CPU time.

First, let's cover getting the run queue size:

• Using vmstat
  As you noted, the r column in vmstat directly represents the run queue size (total runnable processes: running + waiting for CPU). To fetch it quickly:

  vmstat 1 1 | awk 'NR==3 {print $1}'
  

  The 1 1 means we take one sample with a 1-second interval, skipping the header rows.

• Using top
  In the top output, look at the Tasks line—the running value is the run queue size. You can also grab it via command-line mode:

  top -b -n 1 | grep 'Tasks:' | awk '{print $4}'
  

Now, for isolating the number of processes waiting for CPU from vmstat's r value:

The r column counts both processes actively running on CPU and those waiting in the run queue. Since each CPU core can only execute one process at a time, the waiting process count is:

Waiting processes = Total runnable processes (r value) - Number of online CPU cores

If the r value is less than or equal to your core count, that means all runnable processes are actively using CPU—there are no processes waiting.

For example:

• If you have a 4-core system and vmstat shows r=6, then 6-4=2 processes are waiting for CPU.
• If r=3 on the same 4-core system, there are 0 waiting processes.

You can also use pidstat to inspect individual processes, but it's less efficient for a quick count—pidstat -u 1 1 shows CPU usage per process, and processes in R state with low CPU usage are likely waiting for CPU time.

内容的提问来源于stack exchange，提问作者Joey Yi Zhao