-
Notifications
You must be signed in to change notification settings - Fork 1
netdata virtual memory size
You may notice that netdata's virtual memory size, as reported by ps or /proc/pid/status (or even netdata's applications virtual memory chart) is unrealistically high.
For example, it may be reported to be 150+MB, even if the resident memory size is just 25MB. Similar values may be reported for netdata plugins too.
Check this for example: A netdata installation with default settings on Ubuntu 16.04LTS. The top chart is real memory used, while the bottom one is virtual memory:

The system memory allocator allocates virtual memory arenas, per thread running. To my observations, this defaults to 16MB per thread on 64 bit machines. So, if you get the difference between real and virtual memory and divide it by 16MB you will roughly get the number of threads running.
The system does this for speed. Having a separate memory arena for each thread, allows the threads to run in parallel in multi-core systems, without any locks between them.
This behaviour is system specific. For example, the chart above when running netdata on alpine linux (that uses musl instead of glibc) is this:

Since netdata already uses minimal memory allocations while it runs (i.e. it adapts its memory on start, so that while repeatedly collects data it does not do memory allocations), it already instructs the system memory allocator to minimize the memory arenas for each thread. We have also added 2 configuration options to allow you tweak these settings: https://github.com/netdata/netdata/blob/5645b1ee35248d94e6931b64a8688f7f0d865ec6/src/main.c#L410-L418
However, even if we instructed the memory allocator to use just one arena, it seems it allocates an arena per thread.
netdata also supports jemalloc and tcmalloc, however both behave exactly the same to the glibc memory allocator in this aspect.
No, it is not.
Linux reserves real memory (physical RAM) in pages (on x86 machines pages are 4KB each). So even if the system memory allocator is allocating huge amounts of virtual memory, only the 4KB pages that are actually used are reserving physical RAM. The real memory chart on netdata application section, shows the amount of physical memory these pages occupy(it accounts the whole pages, even if parts of them are actually used).
General
- Home
- Why netdata
- Installation
- Installation with docker
- Command Line Options
- Configuration
- Log Files
- Tracing Options
Running Netdata
Special Uses
-
netdata for IoT
lower netdata resource utilization -
high performance netdata
netdata public on the internet
Notes on memory management
-
Memory deduplication
half netdata memory requirements - netdata virtual memory size
Database Replication and Mirroring
- Replication Overview
-
monitoring ephemeral nodes
Use netdata to monitor auto-scaled cloud servers. -
netdata proxies
Streaming netdata metrics between netdata servers.
Backends
archiving netdata collected metrics to a time-series database
-
netdata-backends
graphite,opentsdb,kairosdb,influxdb,elasticsearch,blueflood - netdata with prometheus
- Walk Through: netdata with prometheus and grafana
Health monitoring - Alarms
alarms and alarm notifications in netdata
- Overview
-
Reference
reference for writing alarms -
Examples
simple how-to for writing alarms -
Notifications Configuration
- health API calls
- troubleshooting alarms
Netdata Registry
Monitoring Info
-
Monitoring web servers
The spectacles of a web server log file -
monitoring ephemeral containers
Use netdata to monitor auto-scaled containers. - monitoring systemd services
-
monitoring cgroups
Use netdata to monitor containers and virtual machines. -
monitoring IPMI
Use netdata to monitor enterprise server hardware - Monitoring disks
- Monitoring Go Applications
Netdata Badges
Data Collection
- Add more charts to netdata
- Internal Plugins
- External Plugins
-
statsd
netdata is a fully featured statsd server -
Third Party Plugins
netdata plugins distributed by third parties
Binary Modules
Python Modules
- How to write new module
- apache
- beanstalk
- bind_rndc
- ceph
- couchdb
- cpuidle
- cpufreq
- dns_query_time
- dovecot
- elasticsearch
- exim
- fail2ban
- freeradius
- go_expvar
- haproxy
- hddtemp
- httpcheck
- icecast
- ipfs
- isc_dhcpd
- litespeed
- mdstat
- megacli
- memcached
- mongodb
- mysql
- nginx
- nginx_plus
- nsd
- ntpd
- ovpn_status_log
- phpfpm
- portcheck
- postfix
- postgres
- powerdns
- puppet
- rabbitmq
- redis
- rethinkdbs
- retroshare
- sensors
- spigotmc
- springboot
- squid
- smartd_log
- tomcat
- traefik
- unbound
- varnish
- w1sensor
- web_log
Node.js Modules
BASH Modules
Active BASH Modules
Obsolete BASH Modules
- apache
- cpufreq
- cpu_apps
- exim
- hddtemp
- load_average
- mem_apps
- mysql
- nginx
- phpfpm
- postfix
- sensors
- squid
- tomcat
API Documentation
Web Dashboards
-
Learn how to create dashboards with charts from one or more netdata servers!
Running behind another web server
Package Maintainers
Donations
Blog
-
December, 2016
Linux console tools, fail to report per process CPU usage properly
-
April, 2016
You should install QoS on all your servers (Linux QoS for humans)
Monitor application bandwidth with Linux QoS (Good to do it, anyway)
Monitoring SYNPROXY (Linux TCP Anti-DDoS)
-
March, 2016
Article: Introducing netdata (the design principles of netdata)