Practice it end to end

Take a "sick" machine and work it to ground. Not one symptom, but the whole path: find the bottleneck, harden the system, and prove a copy restores. Do it on a separate VM with a snapshot. You need root on almost every step here.

1. Put load on a clean system. In one terminal start a process that loads the disk and the CPU, for example dd if=/dev/zero of=/var/tmp/load.bin bs=1M count=4000 in a loop, or a load tool like stress-ng. This is your "illness": the system responds slowly and the cause is not yet named.

1. Find the guilty resource by method, not by guessing. Walk the USE checklist over CPU, memory, disk, and network: for each one name utilization, saturation, and errors with vmstat, iostat, free
   Locked

   The rest of this chapter is part of a paid course

   You are reading the open part. Buy the course to read the whole chapter.

   See the course