Monitoring

System monitoring: where to start when the server is slow

The order of checks for "the server is slow": CPU, memory, disk, kernel and network — what shows each resource.

Load average: how many tasks wait for the CPU and the disk

The average number of tasks running or waiting for a CPU (R) or for the kernel (D) over 1, 5 and 15 minutes.

stress-ng — timed load on the CPU, memory and disk

stress-ng puts load on the CPU, memory and disk for a set time — for experiments with load, OOM and disk I/O.

iostat and pidstat — which disk is busy and which process loads it

iostat shows disk load — requests, latency, %util; pidstat shows CPU, memory and disk usage per process.

Memory and swap: free, available and the page cache

How to read free: used and free memory, the page cache and available; swap, a swap file and vm.swappiness.

top — processes and load in real time

The top program refreshes the process list, CPU load, memory and load average every few seconds.

uptime, who, w — who is on the system and the load

Compare load average to the core count (nproc), not an absolute threshold — the three numbers mean nothing without it.

Redis — in-memory key-value store

Cache, queues, and sessions in RAM; KEYS * blocks the server on large databases — use --scan instead.

tcpdump — capturing and analyzing network traffic

Captures packets to diagnose "is traffic even arriving"; without -n it resolves every IP through DNS and slows down.

watch — periodically repeat a command

Repeats a command every N seconds on one screen; without quotes, a pipeline goes to the shell, not the command.

Prometheus — metrics collection

Collecting metrics with a pull model — targets, PromQL, exporters, alerts.

Grafana — dashboards

Visualizing metrics and logs — data sources, dashboards, variables, alerts.