What Linux kernel commands/tools do dynaTrace agents use under the hood to get the system metrics for CPU, Memory etc. Do they use the /proc files?
Need the information as we're experimenting using dynaTrace agents on Linux Containers and want to make sure the agents report the system metrics for the Containers and not the underlying hardware.
Thanks!
Answer by Roman S. ·
Hi,
The host monitoring metrics (that you can see in the Infrastructure view) are using the /proc files. See below for details:
Best, Roman
Metric | Description | Linux | |
CPU Metrics | System Usage | % CPU spent for system/kernel space execution | /proc/stat |
User Usage | % CPU spent for user space execution | /proc/stat | |
Idle | % CPU spent idle | /proc/stat | |
Wait (I/O) | % CPU spent in waiting for I/O | /proc/stat | |
Load | Number of processes running or waiting. "Load" on *nix Platforms, "Processor Queue Length" on Windows Platforms | sysinfo(2) | |
Memory Metrics | Max | /proc/meminfo | |
Available | The available memory is defined as: free+Cached+Buffers | /proc/meminfo | |
Free | |||
Used | /proc/meminfo | ||
Commited | there are two forms of used memory. with and without bufferedram. Used is really used by programs and can not be freed without swapping. Committed is including buffers and caches which are freed if needed. | ||
Cached+Buffers | Can be taken from sysinfo: bufferedram | ||
Swap/Pagefile Metrics | Free | /proc/meminfo | |
Used | /proc/meminfo | ||
Max | /proc/meminfo | ||
Swap Rate/Page faults | major page faults only (requiring disk access) |
/proc/vmstat | |
Network Metrics (per interface) | Rate In | /sys/class/net/*/statistics/rx_bytes | |
Rate Out | /sys/class/net/*/statistics/tx_bytes | ||
Bandwidth | /sys/class/net/*/speed | ||
Disk Metrics(per file system) | Total | /etc/mtab | |
statvfs64(3) | |||
Free | /etc/mtab | ||
statvfs64(3) | |||
Used | /etc/mtab | ||
statvfs64(3) |
JANUARY 15, 3:00 PM GMT / 10:00 AM ET