ContextQMD
Back to Netdata

eBPF Socket

src/collectors/ebpf.plugin/ebpfgo.plugin/README.md

2.11.08.6 KB
Original Source
<!--startmeta custom_edit_url: "https://github.com/netdata/netdata/edit/master/src/collectors/ebpf.plugin/ebpfgo.plugin/integrations/ebpf_socket.md" meta_yaml: "https://github.com/netdata/netdata/edit/master/src/collectors/ebpf.plugin/ebpfgo.plugin/metadata.yaml" sidebar_label: "eBPF Socket" learn_status: "Published" learn_rel_path: "Collecting Metrics/Collectors/Operating Systems" keywords: ['TCP', 'UDP', 'Network monitoring', 'eBPF'] message: "DO NOT EDIT THIS FILE DIRECTLY, IT IS GENERATED BY THE COLLECTOR'S metadata.yaml FILE" endmeta-->

eBPF Socket

Plugin: ebpf-go.plugin Module: socket

Overview

Monitor TCP and UDP function calls, bandwidth, errors, and connection counts. Data is exposed through the on-demand network-protocols function consumed by the network-viewer.

Attach kprobe or fentry/fexit trampolines to the kernel functions tcp_sendmsg, tcp_cleanup_rbuf, tcp_connect (IPv4 and IPv6), tcp_close, udp_sendmsg, and udp_recvmsg. Per-cycle deltas and byte counts are computed in userspace from the BPF map snapshots.

This collector is only supported on the following platforms:

  • Linux

This collector supports collecting metrics from multiple instances of this integration, including remote instances.

The plugin needs setuid because it loads data inside the kernel. Netdata sets the necessary permissions during installation.

Default Behavior

Auto-Detection

The plugin checks kernel compilation flags (CONFIG_KPROBES, CONFIG_BPF, CONFIG_BPF_SYSCALL, CONFIG_BPF_JIT), presence of BTF files, and the configured object flavor to decide which eBPF program will be attached.

Limits

The default configuration for this integration does not impose any limits on data collection.

Performance Impact

This thread adds overhead every time a monitored kernel TCP or UDP function is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology.

Setup

Prerequisites

Compile kernel

Check if your kernel was compiled with necessary options (CONFIG_KPROBES, CONFIG_BPF, CONFIG_BPF_SYSCALL, CONFIG_BPF_JIT) in /proc/config.gz or inside /boot/config file. Some cited names can be different according to the preferences of Linux distributions. When you do not have options set, it is necessary to get the kernel source code from https://kernel.org or a kernel package from your distribution, this last is preferred. The kernel compilation has a well defined pattern, but distributions can deliver their configuration files with different names.

Now follow steps:

  1. Copy the configuration file to /usr/src/linux/.config.
  2. Select the necessary options: make oldconfig
  3. Compile your kernel image: make bzImage
  4. Compile your modules: make modules
  5. Copy your new kernel image for boot loader directory
  6. Install the new modules: make modules_install
  7. Generate an initial ramdisk image (initrd) if it is necessary.
  8. Update your boot loader

Configuration

Options

The socket option in section [ebpf programs] controls whether the module is loaded. Per-module overrides (update every, maps per core, btf path, ebpf object flavor, pid table size, socket monitoring table size, udp connection table size) belong in ebpf.d/socket.conf.

<details open><summary>Config options</summary>
OptionDescriptionDefaultRequired
socketEnable (yes) or disable (no) the socket monitoring module.nono
update everyData collection frequency in seconds.10no
maps per coreWhen enabled (yes), the plugin allocates one BPF hash table per CPU core instead of a single shared table. Improves throughput on multi-core systems; ignored on kernels older than 4.15.yesno
btf pathPath to the directory containing BTF files used for CO-RE (Compile Once, Run Everywhere) loading./sys/kernel/btfno
ebpf object flavorSelect the BPF object flavor. Available values are arena, buffer ring (also accepted as buffer), and legacy (also accepted as tracing). The collector falls back to the next available flavor when the requested one is not supported by the running kernel.bufferno
pid table sizeMaximum number of PIDs tracked in the per-PID SHM segment shared with apps.plugin and cgroups.plugin. Increase on hosts with more than 32768 concurrent monitored processes.32768no
socket monitoring table sizeMaximum number of connections tracked in the tbl_nd_socket BPF hash map (per-PID TCP/UDP bandwidth). Increase on hosts with many concurrent connections; each entry consumes kernel BPF map memory.16384no
udp connection table sizeMaximum number of UDP connections tracked in the tbl_nv_udp BPF hash map. Increase on hosts with high UDP connection rates.4096no
</details>

via File

The configuration file name for this integration is ebpf.d.conf.

You can edit the configuration file using the edit-config script from the Netdata config directory.

bash
cd /etc/netdata 2>/dev/null || cd /opt/netdata/etc/netdata
sudo ./edit-config ebpf.d.conf
Examples

There are no configuration examples.

Alerts

There are no alerts configured by default for this integration.

Metrics

Metrics grouped by scope.

The scope defines the instance that the metric belongs to. An instance is uniquely identified by a set of labels.

This module attaches to kernel TCP/UDP functions and publishes host-level TCP/UDP activity through the on-demand network-protocols function (FUNCTIONGLOBAL) consumed by the network-viewer. Per-cgroup and per-service socket activity is exported through the eBPFGo shared-memory path consumed by cgroups.plugin.

Migration note — Per-host and per-app charts published by prior versions of ebpf.plugin are not republished. The following are permanently removed; use the network-protocols function in network-viewer as a replacement.

Removed per-host charts: ip.inbound_conn, ip.tcp_outbound_conn, ip.tcp_functions, ip.total_tcp_bandwidth, ip.tcp_error, ip.tcp_retransmit, ip.udp_functions, ip.total_udp_bandwidth, ip.udp_error.

Removed per-app charts: app.ebpf_call_tcp_v4_connection, app.ebpf_call_tcp_v6_connection, app.ebpf_sock_total_bandwidth, app.ebpf_call_tcp_sendmsg, app.ebpf_call_tcp_cleanup_rbuf, app.ebpf_call_tcp_retransmit, app.ebpf_call_udp_sendmsg, app.ebpf_call_udp_recvmsg.

Configuration changes from network.conf — The network.conf file has been replaced by socket.conf. The following defaults changed silently:

  • maps per core changed from no to yes. Per-core BPF maps use more memory but reduce contention on multi-core systems. To restore the previous behaviour add maps per core = no to socket.conf.
  • ebpf co-re tracing changed from probe (kprobe/legacy BPF) to trampoline (auto-detected). To force kprobes add ebpf co-re tracing = probe to socket.conf.
  • The [network connections] and [service name] sections from network.conf have no equivalent in socket.conf and are not read; any per-port service name overrides are lost.

network-protocols function columns

The network-protocols function returns a table with one row per transport protocol and the following columns:

ColumnDescription
TransportProtocol name (TCP or UDP)
FamilyIP family (IPv4+IPv6 combined)
ReceivedSegments or datagrams received per second
SentSegments or datagrams sent per second
ErrorsSend/receive errors per second
ConnActiveActive connections opened per second
ConnEstablishedCurrently established connections
ConnPassivePassive connections opened per second
ConnResetReset connections per second
SegsTotalTotal segments per second
SegsRetransmittedRetransmitted segments per second
DatagramsNoPortDatagrams with no destination port per second

Per eBPF Socket instance

All data is delivered through the on-demand network-protocols function consumed by network-viewer; no traditional Netdata charts are published for this scope.

This scope has no labels.

Metrics:

MetricDimensionsUnit