0

I got bare-metal server with 32cores/64thrds CPU and 64GB RAM. Ubuntu 18.04 on-board. Running pgsql (low load), ~250 docker containers with python for network working. htop show me LA 120 (1/5/15 mins), but i can not find the reason.

  • htop show ~50% load of all cores
  • Got 12GB free RAM
  • vnstat -l show 25% load on network adapter (in both sides summary)
  • top show wa (iowait) 0.0
  • iostat show nothing writen on HDDs, iotop sometimes show processes with writing 10-20Kb/sec on one second.
  • atop do not show red lines. All gray (good).
  • "ps -eo stat | grep -c D" and "ps -eo stat | grep -c D" show "1"

I think i checked all - CPU, RAM, HDD, network. I understand the root of big LA - containers with python. But i want find a bottleneck for optimisation. Where can i see a reason of highly LA more?

UPD: Some additional. I think, bottleneck somewhere in CPU because if i use pypy in conrainers LA go down, if use simple pyhton - LA grow up. But i can not understand the since because htop always show CPU load <100%.

Anton K.
  • 11
  • 3

1 Answers1

0

Linux load average is the number of processes currently (actively) running or waiting for execution.

You have 250 Docker containers plus system processes and your monitoring tools. This is simply a huge amount of processes. It depends a bit what's running inside your containers and how often processes are waking up to do something. If you reduce the number of Docker containers by 50% and the load average is going down by 50% as well then you have the culprit.

C.B.
  • 21
  • 5