Just use http module

[caub]

I saw you kinda reimplemented http in dockerSocket.js

you could just use http module, example in this lib I made

fetchu({ path: '/v1.39/services', socketPath: '/var/run/docker.sock' })

from discussion in node-fetch/node-fetch#336 (comment)

[sebhildebrandt]

@caub thank you for your comment! Will think how this can improve the library. Generally I want to keep dependencies very low ... and as sockets work fine so far, I will think of how I can/should use http module ... closing it for now.

[caub]

@sebhildebrandt Thanks, fair enough

One quick question, could you comment on how cpuPercent is calculated in your lib, and below:

const dockerStats = async containerId => {
  const { name, cpu_stats, precpu_stats, memory_stats } = await fetch(`http://v1.39/containers/${containerId}/stats?stream=false`, { socketPath: '/var/run/docker.sock' });
  const cpuPercent = (cpu_stats.cpu_usage.total_usage - precpu_stats.cpu_usage.total_usage) / (cpu_stats.system_cpu_usage - precpu_stats.system_cpu_usage) * cpu_stats.cpu_usage.percpu_usage.length * 100;
  return {id, name, cpuPercent, memPercent: memory_stats.usage * 100 / memory_stats.limit};
}

Is this the percentage of increase of CPU? (if precpu_stats stands for previouscpu_stats). I doubt since it's constantly positive while the process is stable
What does it measure so?

[sebhildebrandt]

So this is a longer story. First: yes, the calculation is based on previous stats. I actually did this wrong but then I based my calculation on the way how docker does it in his own code for calculating stats (actually to get in sync with docker stats). Took me some while to get there, but now I think this should be correct. In cpu_stats.cpu_usage.total_usage, ... there are no percent values so the result must be always positive. Any doubts?

[caub]

Thanks, it looks weird, but yea, good idea I'll look at docker CLI source