@cryptotaxi247 / netdata / commits / ae2f6b6c8

docs: fix eBPF performance overhead unit from ms to ns in metadata.yaml (#22277)

* docs: fix eBPF performance overhead unit from ms to ns in metadata.yaml The performance_impact descriptions for all 10 eBPF module entries (filedescriptor, cachestat, sync, mdflush, swap, socket, dcstat, shm, mount, vfs) incorrectly stated '90-200ms per call' for overhead on kernels without BTF. The authoritative README.md states kprobes add 'around 100ns overhead for each function call' (line 440), making nanoseconds the correct unit. Changed '90-200ms' to '90-200ns' to align with the actual overhead measurement. * docs: fix eBPF socket metric dimension names in metadata.yaml Fix two dimension name discrepancies between metadata.yaml and the actual source code (ebpf_socket.c): - ip.tcp_retransmit: correct dimension name from 'retransmited' to 'retransmitted' to match the dimension name defined in socket_dimension_names[] at ebpf_socket.c:21. - ip.inbound_conn: correct dimension name from 'connection_tcp' to 'connected_tcp' and add the missing 'connected_udp' dimension, as both are used in the chart creation at ebpf_socket.c:1051-1052. --------- Co-authored-by: nedi-app[bot] <nedi-app[bot]@users.noreply.github.com>

nedi-app[bot] committed May 4, 2026 at 07:42 UTC ae2f6b6c8999b34845ea7374f07aca8bc87b4e9d

1 file changed +13 -12

src/collectors/ebpf.plugin/metadata.yaml

+13 -12

@@ -42,7 +42,7 @@ modules:

42 limits:

43 description: ""

44 performance_impact:

45 - description: "Depending of kernel version and frequency that files are open and close, this thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

45 + description: "Depending of kernel version and frequency that files are open and close, this thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

46 setup:

47 prerequisites:

48 list:

@@ -731,7 +731,7 @@ modules:

731 limits:

732 description: ""

733 performance_impact:

734 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

734 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

735 setup:

736 prerequisites:

737 list:

@@ -959,7 +959,7 @@ modules:

959 limits:

960 description: ""

961 performance_impact:

962 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

962 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

963 setup:

964 prerequisites:

965 list:

@@ -1139,7 +1139,7 @@ modules:

1139 limits:

1140 description: ""

1141 performance_impact:

1142 - description: "This thread will add overhead every time that `md_flush_request` is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

1142 + description: "This thread will add overhead every time that `md_flush_request` is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

1143 setup:

1144 prerequisites:

1145 list:

@@ -1248,7 +1248,7 @@ modules:

1248 limits:

1249 description: ""

1250 performance_impact:

1251 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

1251 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

1252 setup:

1253 prerequisites:

1254 list:

@@ -1558,7 +1558,7 @@ modules:

1558 limits:

1559 description: ""

1560 performance_impact:

1561 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

1561 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

1562 setup:

1563 prerequisites:

1564 list:

@@ -1661,7 +1661,8 @@ modules:

1661 unit: "connections/s"

1662 chart_type: line

1663 dimensions:

1664 - - name: connection_tcp

1664 + - name: connected_tcp

1665 + - name: connected_udp

1666 - name: ip.tcp_outbound_conn

1667 description: TCP outbound connections.

1668 unit: "connections/s"

@@ -1695,7 +1696,7 @@ modules:

1696 unit: "calls/s"

1697 chart_type: line

1698 dimensions:

1698 - - name: retransmited

1699 + - name: retransmitted

1700 - name: ip.udp_functions

1701 description: UDP calls

1702 unit: "calls/s"

@@ -1914,7 +1915,7 @@ modules:

1915 limits:

1916 description: ""

1917 performance_impact:

1917 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

1918 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

1919 setup:

1920 prerequisites:

1921 list:

@@ -2297,7 +2298,7 @@ modules:

2298 limits:

2299 description: ""

2300 performance_impact:

2300 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

2301 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

2302 setup:

2303 prerequisites:

2304 list:

@@ -2634,7 +2635,7 @@ modules:

2635 limits:

2636 description: ""

2637 performance_impact:

2637 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

2638 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

2639 setup:

2640 prerequisites:

2641 list:

@@ -2763,7 +2764,7 @@ modules:

2764 limits:

2765 description: ""

2766 performance_impact:

2766 - description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ms per call on kernels that do not have BTF technology."

2767 + description: "This thread will add overhead every time that an internal kernel function monitored by this thread is called. The estimated additional period of time is between 90-200ns per call on kernels that do not have BTF technology."

2768 setup:

2769 prerequisites:

2770 list: