Pure python scripts for performance measurements based on linux /proc and /sys filesystems

This is a set of tools written as pure python scripts without any external dependencies. Just copy script to target server and run it with python interpreter (both 2.x and 3.x is OK). Check source code to understand what exactly is measured.

Get user and system CPU time consumed by target process

Use linux-cpu-usage script to get CPU time counters from /proc/[pid]/stat. By default it just prints current utime, stime, cutime and cstime (see man 5 proc) as comma-separated float values in seconds.

You can store script output before some work is done by process and run script afterwards passing stored value as --delta-from argument to get counters change between two measurements.

Use --total argument to get total CPU usage in seconds or --efficiency-for-amount to get CPU efficiency for given amount of work (i.e. given amount divided by total CPU usage time in seconds).

Target process could be identified ether directly by it's PID, or using regular expression for command line.

See also ./linux-cpu-usage --help.

Usage examples:

$ ./linux-cpu-usage -c sshd
0.16,0.18,0.00,0.00

$ start=$(./linux-cpu-usage -c Xorg); \
    sleep 10s; \
    ./linux-cpu-usage -c Xorg -d $start
0.13,0.09,0.00,0.00

$ ./linux-cpu-usage -p 2740 -t
15742.11

Measure network counters change rate

The linux-net-rate is a simple tool measuring rate of change for network interface counters in /sys/class/net/{interface}/statistics.

Usage example:

./linux-net-rate -i 0.1 -c 5 bond0:tx_bytes:8 bond0:rx_bytes:8 bond0:tx_errors
time         bond0:tx_bytes:8     bond0:rx_bytes:8      bond0:tx_errors
22:03:53               8.694G             665.331M               0.000
22:03:53               8.356G             707.296M               0.000
22:03:53               7.760G               1.240G               0.000
22:03:53               7.822G             406.817M               0.000
22:03:53               8.272G             224.498M               0.000