428d925be · cryptotaxi247/netdata-1

@@ -8,51 +8,405 @@ sidebar_label: "eBPF"

8

9 # eBPF monitoring with Netdata

10

11 -eBPF consists of a wide toolchain that ultimately outputs a set of bytecode that will run inside the eBPF virtual

12 -machine (VM) which lives inside the Linux kernel. The program in particular is executed in response to a [tracepoint

13 -or kprobe](#probes-and-tracepoints) activation.

11 +The Netdata Agent provides many [eBPF](https://ebpf.io/what-is-ebpf/) programs to help you troubleshoot and debug how applications interact with the Linux kernel. The `ebpf.plugin` uses [tracepoints, trampoline, and2 kprobes](#data-collection) to collect a wide array of high value data about the host that would otherwise be impossible to capture.

12

15 -Netdata has written many eBPF programs, which, when compiled and integrated into the Netdata Agent, are able to collect

16 -a wide array of data about the host that would otherwise be impossible. The data eBPF programs can collect is truly unique,

17 -which gives the Netdata Agent access to data that is high value but normally hard to capture.

13 +> ❗ eBPF monitoring only works on Linux systems and with specific Linux kernels, including all kernels newer than `4.11.0`, and all kernels on CentOS 7.6 or later. For kernels older than `4.11.0`, improved support is in active development.

14

19 -eBPF monitoring can help you troubleshoot and debug how applications interact with the Linux kernel. See

20 -our [guide on troubleshooting apps with eBPF metrics](/docs/guides/troubleshoot/monitor-debug-applications-ebpf.md) for

21 -configuration and troubleshooting tips.

15 +This document provides comprehensive details about the `ebpf.plugin`.

16 +For hands-on configuration and troubleshooting tips see our [tutorial on troubleshooting apps with eBPF metrics](/docs/guides/troubleshoot/monitor-debug-applications-ebpf.md).

17

18 <figure>

19 <img src="https://user-images.githubusercontent.com/1153921/74746434-ad6a1e00-5222-11ea-858a-a7882617ae02.png" alt="An example of VFS charts, made possible by the eBPF collector plugin" />

25 - <figcaption>An example of VFS charts made possible by the eBPF collector plugin.</figcaption>

20 + <figcaption>An example of virtual file system (VFS) charts made possible by the eBPF collector plugin.</figcaption>

21 </figure>

22

28 -## Probes and Tracepoints

23 +<a id="data-collection"> </a>

24

30 -The following two features from the Linux kernel are used by Netdata to run eBPF programs:

25 +## How Netdata collects data using probes and tracepoints

26

32 -- Kprobes and return probes (kretprobe): Probes can insert virtually into any kernel instruction. When eBPF runs in

33 - `entry` mode, it attaches only `kprobes` for internal functions monitoring calls and some arguments every time a

34 - function is called. The user can also change configuration to use [`return`](#global) mode, and this will allow users

35 - to monitor return from these functions and detect possible failures.

36 -- Tracepoints are hooks to call specific functions. Tracepoints are more stable than `kprobes` and are preferred when

37 - both options are available.

27 +Netdata uses the following features from the Linux kernel to run eBPF programs:

28

39 -In each case, wherever a normal kprobe, kretprobe, or tracepoint would have run its hook function, an eBPF program is

40 -run instead, performing various collection logic before letting the kernel continue its normal control flow.

29 +- Tracepoints are hooks to call specific functions. Tracepoints are more stable than `kprobes` and are preferred when

30 + both options are available.

31 +- Trampolines are bridges between kernel functions, and BPF programs. Netdata uses them by default whenever available.

32 +- Kprobes and return probes (`kretprobe`): Probes can insert virtually into any kernel instruction. When eBPF runs in `entry` mode, it attaches only `kprobes` for internal functions monitoring calls and some arguments every time a function is called. The user can also change configuration to use [`return`](#global) mode, and this will allow users to monitor return from these functions and detect possible failures.

33

42 -There are more methods by which eBPF programs can be triggered but which are not currently supported, such as via uprobes

43 -which allow hooking into arbitrary user-space functions in a similar manner to kprobes.

34 +In each case, wherever a normal kprobe, kretprobe, or tracepoint would have run its hook function, an eBPF program is run instead, performing various collection logic before letting the kernel continue its normal control flow.

35

45 -## Manually enable the collector on Linux

36 +There are more methods to trigger eBPF programs, such as uprobes, but currently are not supported.

37

47 -**The eBPF collector is installed and enabled by default on most new installations of the Agent**. The eBPF collector

48 -does not currently work with [static build installations](/packaging/installer/methods/kickstart.md#static-builds) for kernels older

49 -than `4.11`, but improved support is in active development.

38 +## Configuring ebpf.plugin

39

51 -eBPF monitoring only works on Linux systems and with specific Linux kernels, including all kernels newer than `4.11.0`,

52 -and all kernels on CentOS 7.6 or later.

40 +The eBPF collector is installed and enabled by default on most new installations of the Agent.

41 +If your Agent is v1.22 or older, you may to enable the collector yourself.

42

54 -If your Agent is v1.22 or older, you may to enable the collector yourself. See the [configuration](#configuration)

55 -section for details.

43 +### Enable the eBPF collector

44 +

45 +To enable or disable the entire eBPF collector:

46 +

47 +1. Navigate to the [Netdata config directory](/docs/configure/nodes.md#the-netdata-config-directory).

48 + ```bash

49 + cd /etc/netdata

50 + ```

51 +

52 +2. Use the [`edit-config`](/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit `netdata.conf`.

53 +

54 + ```bash

55 + ./edit-config netdata.conf

56 + ```

57 +

58 +3. Enable the collector by scrolling down to the `[plugins]` section. Uncomment the line `ebpf` (not

59 + `ebpf_process`) and set it to `yes`.

60 +

61 + ```conf

62 + [plugins]

63 + ebpf = yes

64 + ```

65 +

66 +### Configure the eBPF collector

67 +

68 +You can configure the eBPF collector's behavior to fine-tune which metrics you receive and [optimize performance]\(#performance opimization).

69 +

70 +To edit the `ebpf.d.conf`:

71 +

72 +1. Navigate to the [Netdata config directory](/docs/configure/nodes.md#the-netdata-config-directory).

73 + ```bash

74 + cd /etc/netdata

75 + ```

76 +2. Use the [`edit-config`](/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit [`ebpf.d.conf`](https://github.com/netdata/netdata/blob/master/collectors/ebpf.plugin/ebpf.d.conf).

77 +

78 + ```bash

79 + ./edit-config ebpf.d.conf

80 + ```

81 +

82 + You can now edit the behavior of the eBPF collector. The following sections describe each configuration option in detail.

83 +

84 +### `[global]` configuration options

85 +

86 +The `[global]` section defines settings for the whole eBPF collector.

87 +

88 +#### eBPF load mode

89 +

90 +The collector uses two different eBPF programs. These programs rely on the same functions inside the kernel, but they

91 +monitor, process, and display different kinds of information.

92 +

93 +By default, this plugin uses the `entry` mode. Changing this mode can create significant overhead on your operating

94 +system, but also offer valuable information if you are developing or debugging software. The `ebpf load mode` option

95 +accepts the following values:

96 +

97 +- `entry`: This is the default mode. In this mode, the eBPF collector only monitors calls for the functions described in

98 + the sections above, and does not show charts related to errors.

99 +- `return`: In the `return` mode, the eBPF collector monitors the same kernel functions as `entry`, but also creates new

100 + charts for the return of these functions, such as errors. Monitoring function returns can help in debugging software,

101 + such as failing to close file descriptors or creating zombie processes.

102 +- `update every`: Number of seconds used for eBPF to send data for Netdata.

103 +- `pid table size`: Defines the maximum number of PIDs stored inside the application hash table.

104 +

105 +#### Integration with `apps.plugin`

106 +

107 +The eBPF collector also creates charts for each running application through an integration with the

108 +[`apps.plugin`](/collectors/apps.plugin/README.md). This integration helps you understand how specific applications

109 +interact with the Linux kernel.

110 +

111 +If you want to _disable_ the integration with `apps.plugin` along with the above charts, change the setting `apps` to

112 +`no`.

113 +

114 +```conf

115 +[global]

116 + apps = yes

117 +```

118 +

119 +When the integration is enabled, eBPF collector allocates memory for each process running. The total allocated memory

120 +has direct relationship with the kernel version. When the eBPF plugin is running on kernels newer than `4.15`, it uses

121 +per-cpu maps to speed up the update of hash tables. This also implies storing data for the same PID for each processor

122 +it runs.

123 +

124 +#### Integration with `cgroups.plugin`

125 +

126 +The eBPF collector also creates charts for each cgroup through an integration with the

127 +[`cgroups.plugin`](/collectors/cgroups.plugin/README.md). This integration helps you understand how a specific cgroup

128 +interacts with the Linux kernel.

129 +

130 +The integration with `cgroups.plugin` is disabled by default to avoid creating overhead on your system. If you want to

131 +_enable_ the integration with `cgroups.plugin`, change the `cgroups` setting to `yes`.

132 +

133 +```conf

134 +[global]

135 + cgroups = yes

136 +```

137 +

138 +If you do not need to monitor specific metrics for your `cgroups`, you can enable `cgroups` inside

139 +`ebpf.d.conf`, and then disable the plugin for a specific `thread` by following the steps in the

140 +[Configuration](#configuration) section.

141 +

142 +#### Integration Dashboard Elements

143 +

144 +When an integration is enabled, your dashboard will also show the following cgroups and apps charts using low-level

145 +Linux metrics:

146 +

147 +> Note: The parenthetical accompanying each bulleted item provides the chart name.

148 +

149 +- mem

150 + - Number of processes killed due out of memory. (`oomkills`)

151 +- process

152 + - Number of processes created with `do_fork`. (`process_create`)

153 + - Number of threads created with `do_fork` or `clone (2)`, depending on your system's kernel

154 + version. (`thread_create`)

155 + - Number of times that a process called `do_exit`. (`task_exit`)

156 + - Number of times that a process called `release_task`. (`task_close`)

157 + - Number of times that an error happened to create thread or process. (`task_error`)

158 +- swap

159 + - Number of calls to `swap_readpage`. (`swap_read_call`)

160 + - Number of calls to `swap_writepage`. (`swap_write_call`)

161 +- network

162 + - Number of outbound connections using TCP/IPv4. (`outbound_conn_ipv4`)

163 + - Number of outbound connections using TCP/IPv6. (`outbound_conn_ipv6`)

164 + - Number of bytes sent. (`total_bandwidth_sent`)

165 + - Number of bytes received. (`total_bandwidth_recv`)

166 + - Number of calls to `tcp_sendmsg`. (`bandwidth_tcp_send`)

167 + - Number of calls to `tcp_cleanup_rbuf`. (`bandwidth_tcp_recv`)

168 + - Number of calls to `tcp_retransmit_skb`. (`bandwidth_tcp_retransmit`)

169 + - Number of calls to `udp_sendmsg`. (`bandwidth_udp_send`)

170 + - Number of calls to `udp_recvmsg`. (`bandwidth_udp_recv`)

171 +- file access

172 + - Number of calls to open files. (`file_open`)

173 + - Number of calls to open files that returned errors. (`open_error`)

174 + - Number of files closed. (`file_closed`)

175 + - Number of calls to close files that returned errors. (`file_error_closed`)

176 +- vfs

177 + - Number of calls to `vfs_unlink`. (`file_deleted`)

178 + - Number of calls to `vfs_write`. (`vfs_write_call`)

179 + - Number of calls to write a file that returned errors. (`vfs_write_error`)

180 + - Number of calls to `vfs_read`. (`vfs_read_call`)

181 + - - Number of calls to read a file that returned errors. (`vfs_read_error`)

182 + - Number of bytes written with `vfs_write`. (`vfs_write_bytes`)

183 + - Number of bytes read with `vfs_read`. (`vfs_read_bytes`)

184 + - Number of calls to `vfs_fsync`. (`vfs_fsync`)

185 + - Number of calls to sync file that returned errors. (`vfs_fsync_error`)

186 + - Number of calls to `vfs_open`. (`vfs_open`)

187 + - Number of calls to open file that returned errors. (`vfs_open_error`)

188 + - Number of calls to `vfs_create`. (`vfs_create`)

189 + - Number of calls to open file that returned errors. (`vfs_create_error`)

190 +- page cache

191 + - Ratio of pages accessed. (`cachestat_ratio`)

192 + - Number of modified pages ("dirty"). (`cachestat_dirties`)

193 + - Number of accessed pages. (`cachestat_hits`)

194 + - Number of pages brought from disk. (`cachestat_misses`)

195 +- directory cache

196 + - Ratio of files available in directory cache. (`dc_hit_ratio`)

197 + - Number of files accessed. (`dc_reference`)

198 + - Number of files accessed that were not in cache. (`dc_not_cache`)

199 + - Number of files not found. (`dc_not_found`)

200 +- ipc shm

201 + - Number of calls to `shm_get`. (`shmget_call`)

202 + - Number of calls to `shm_at`. (`shmat_call`)

203 + - Number of calls to `shm_dt`. (`shmdt_call`)

204 + - Number of calls to `shm_ctl`. (`shmctl_call`)

205 +

206 +### `[ebpf programs]` configuration options

207 +

208 +The eBPF collector enables and runs the following eBPF programs by default:

209 +

210 +- `fd` : This eBPF program creates charts that show information about calls to open files.

211 +- `mount`: This eBPF program creates charts that show calls to syscalls mount(2) and umount(2).

212 +- `shm`: This eBPF program creates charts that show calls to syscalls shmget(2), shmat(2), shmdt(2) and shmctl(2).

213 +- `sync`: Monitor calls to syscalls sync(2), fsync(2), fdatasync(2), syncfs(2), msync(2), and sync_file_range(2).

214 +- `network viewer`: This eBPF program creates charts with information about `TCP` and `UDP` functions, including the

215 + bandwidth consumed by each.

216 +- `vfs`: This eBPF program creates charts that show information about VFS (Virtual File System) functions.

217 +- `process`: This eBPF program creates charts that show information about process life. When in `return` mode, it also

218 + creates charts showing errors when these operations are executed.

219 +- `hardirq`: This eBPF program creates charts that show information about time spent servicing individual hardware

220 + interrupt requests (hard IRQs).

221 +- `softirq`: This eBPF program creates charts that show information about time spent servicing individual software

222 + interrupt requests (soft IRQs).

223 +- `oomkill`: This eBPF program creates a chart that shows OOM kills for all applications recognized via

224 + the `apps.plugin` integration. Note that this program will show application charts regardless of whether apps

225 + integration is turned on or off.

226 +

227 +You can also enable the following eBPF programs:

228 +

229 +- `cachestat`: Netdata's eBPF data collector creates charts about the memory page cache. When the integration with

230 + [`apps.plugin`](/collectors/apps.plugin/README.md) is enabled, this collector creates charts for the whole host _and_

231 + for each application.

232 +- `dcstat` : This eBPF program creates charts that show information about file access using directory cache. It appends

233 + `kprobes` for `lookup_fast()` and `d_lookup()` to identify if files are inside directory cache, outside and files are

234 + not found.

235 +- `disk` : This eBPF program creates charts that show information about disk latency independent of filesystem.

236 +- `filesystem` : This eBPF program creates charts that show information about some filesystem latency.

237 +- `swap` : This eBPF program creates charts that show information about swap access.

238 +- `mdflush`: This eBPF program creates charts that show information about

239 + multi-device software flushes.

240 +

241 +### Configuring eBPF threads

242 +

243 +You can configure each thread of the eBPF data collector. This allows you to overwrite global options defined in `/etc/netdata/ebpf.d.conf` and configure specific options for each thread.

244 +

245 +To configure an eBPF thread:

246 +

247 +1. Navigate to the [Netdata config directory](/docs/configure/nodes.md#the-netdata-config-directory).

248 + ```bash

249 + cd /etc/netdata

250 + ```

251 +2. Use the [`edit-config`](/docs/configure/nodes.md#use-edit-config-to-edit-configuration-files) script to edit a thread configuration file. The following configuration files are available:

252 +

253 + - `network.conf`: Configuration for the [`network` thread](#network-configuration). This config file overwrites the global options and also

254 + lets you specify which network the eBPF collector monitors.

255 + - `process.conf`: Configuration for the [`process` thread](#sync-configuration).

256 + - `cachestat.conf`: Configuration for the `cachestat` thread(#filesystem-configuration).

257 + - `dcstat.conf`: Configuration for the `dcstat` thread.

258 + - `disk.conf`: Configuration for the `disk` thread.

259 + - `fd.conf`: Configuration for the `file descriptor` thread.

260 + - `filesystem.conf`: Configuration for the `filesystem` thread.

261 + - `hardirq.conf`: Configuration for the `hardirq` thread.

262 + - `softirq.conf`: Configuration for the `softirq` thread.

263 + - `sync.conf`: Configuration for the `sync` thread.

264 + - `vfs.conf`: Configuration for the `vfs` thread.

265 +

266 + ```bash

267 + ./edit-config FILE.conf

268 + ```

269 +

270 +### Network configuration

271 +

272 +The network configuration has specific options to configure which network(s) the eBPF collector monitors. These options

273 +are divided in the following sections:

274 +

275 +#### `[network connections]`

276 +

277 +You can configure the information shown on `outbound` and `inbound` charts with the settings in this section.

278 +

279 +```conf

280 +[network connections]

281 + maximum dimensions = 500

282 + resolve hostname ips = no

283 + ports = 1-1024 !145 !domain

284 + hostnames = !example.com

285 + ips = !127.0.0.1/8 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 fc00::/7

286 +```

287 +

288 +When you define a `ports` setting, Netdata will collect network metrics for that specific port. For example, if you

289 +write `ports = 19999`, Netdata will collect only connections for itself. The `hostnames` setting accepts

290 +[simple patterns](/libnetdata/simple_pattern/README.md). The `ports`, and `ips` settings accept negation (`!`) to deny

291 +specific values or asterisk alone to define all values.

292 +

293 +In the above example, Netdata will collect metrics for all ports between 1 and 443, with the exception of 53 (domain)

294 +and 145.

295 +

296 +The following options are available:

297 +

298 +- `ports`: Define the destination ports for Netdata to monitor.

299 +- `hostnames`: The list of hostnames that can be resolved to an IP address.

300 +- `ips`: The IP or range of IPs that you want to monitor. You can use IPv4 or IPv6 addresses, use dashes to define a

301 + range of IPs, or use CIDR values. By default, only data for private IP addresses is collected, but this can

302 + be changed with the `ips` setting.

303 +

304 +By default, Netdata displays up to 500 dimensions on network connection charts. If there are more possible dimensions,

305 +they will be bundled into the `other` dimension. You can increase the number of shown dimensions by changing

306 +the `maximum dimensions` setting.

307 +

308 +The dimensions for the traffic charts are created using the destination IPs of the sockets by default. This can be

309 +changed setting `resolve hostname ips = yes` and restarting Netdata, after this Netdata will create dimensions using

310 +the `hostnames` every time that is possible to resolve IPs to their hostnames.

311 +

312 +#### `[service name]`

313 +

314 +Netdata uses the list of services in `/etc/services` to plot network connection charts. If this file does not contain

315 +the name for a particular service you use in your infrastructure, you will need to add it to the `[service name]`

316 +section.

317 +

318 +For example, Netdata's default port (`19999`) is not listed in `/etc/services`. To associate that port with the Netdata

319 +service in network connection charts, and thus see the name of the service instead of its port, define it:

320 +

321 +```conf

322 +[service name]

323 + 19999 = Netdata

324 +```

325 +

326 +### Sync configuration

327 +

328 +The sync configuration has specific options to disable monitoring for syscalls. All syscalls are monitored by default.

329 +

330 +```conf

331 +[syscalls]

332 + sync = yes

333 + msync = yes

334 + fsync = yes

335 + fdatasync = yes

336 + syncfs = yes

337 + sync_file_range = yes

338 +```

339 +

340 +### Filesystem configuration

341 +

342 +The filesystem configuration has specific options to disable monitoring for filesystems; by default, all filesystems are

343 +monitored.

344 +

345 +```conf

346 +[filesystem]

347 + btrfsdist = yes

348 + ext4dist = yes

349 + nfsdist = yes

350 + xfsdist = yes

351 + zfsdist = yes

352 +```

353 +

354 +The ebpf program `nfsdist` monitors only `nfs` mount points.

355 +

356 +## Troubleshooting

357 +

358 +If the eBPF collector does not work, you can troubleshoot it by running the `ebpf.plugin` command and investigating its

359 +output.

360 +

361 +```bash

362 +cd /usr/libexec/netdata/plugins.d/

363 +sudo su -s /bin/bash ./ebpf.plugin

364 +```

365 +

366 +You can also use `grep` to search the Agent's `error.log` for messages related to eBPF monitoring.

367 +

368 +```bash

369 +grep -i ebpf /var/log/netdata/error.log

370 +```

371 +

372 +### Confirm kernel compatibility

373 +

374 +The eBPF collector only works on Linux systems and with specific Linux kernels. We support all kernels more recent than

375 +`4.11.0`, and all kernels on CentOS 7.6 or later.

376 +

377 +You can run our helper script to determine whether your system can support eBPF monitoring. If it returns no output, your system is ready to compile and run the eBPF collector.

378 +

379 +```bash

380 +curl -sSL https://raw.githubusercontent.com/netdata/kernel-collector/master/tools/check-kernel-config.sh | sudo bash

381 +```

382 +

383 +

384 +If you see a warning about a missing kernel

385 +configuration (`KPROBES KPROBES_ON_FTRACE HAVE_KPROBES BPF BPF_SYSCALL BPF_JIT`), you will need to recompile your kernel

386 +to support this configuration. The process of recompiling Linux kernels varies based on your distribution and version.

387 +Read the documentation for your system's distribution to learn more about the specific workflow for recompiling the

388 +kernel, ensuring that you set all the necessary

389 +

390 +- [Ubuntu](https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel)

391 +- [Debian](https://kernel-team.pages.debian.net/kernel-handbook/ch-common-tasks.html#s-common-official)

392 +- [Fedora](https://fedoraproject.org/wiki/Building_a_custom_kernel)

393 +- [CentOS](https://wiki.centos.org/HowTos/Custom_Kernel)

394 +- [Arch Linux](https://wiki.archlinux.org/index.php/Kernel/Traditional_compilation)

395 +- [Slackware](https://docs.slackware.com/howtos:slackware_admin:kernelbuilding)

396 +

397 +### Mount `debugfs` and `tracefs`

398 +

399 +The eBPF collector also requires both the `tracefs` and `debugfs` filesystems. Try mounting the `tracefs` and `debugfs`

400 +filesystems using the commands below:

401 +

402 +```bash

403 +sudo mount -t debugfs nodev /sys/kernel/debug

404 +sudo mount -t tracefs nodev /sys/kernel/tracing

405 +```

406 +

407 +If they are already mounted, you will see an error. You can also configure your system's `/etc/fstab` configuration to

408 +mount these filesystems on startup. More information can be found in

409 +the [ftrace documentation](https://www.kernel.org/doc/Documentation/trace/ftrace.txt).

410

411 ## Charts

412

@@ -64,8 +418,7 @@ collected in the previous and current seconds.

418

419 ### System overview

420

67 -Not all charts within the System Overview menu are enabled by default, because they add around 100ns overhead for each

68 -function call, this number is small for a human perspective, but the functions are called many times creating an impact

421 +Not all charts within the System Overview menu are enabled by default. Charts that rely on `kprobes` are disabled by default because they add around 100ns overhead for each function call. This is a small number from a human's perspective, but the functions are called many times and create an impact

422 on host. See the [configuration](#configuration) section for details about how to enable them.

423

424 #### Processes

@@ -74,12 +427,12 @@ Internally, the Linux kernel treats both processes and threads as `tasks`. To cr

427 system calls: `fork(2)`, `vfork(2)`, and `clone(2)`. To generate this chart, the eBPF

428 collector uses the following `tracepoints` and `kprobe`:

429

77 -- `sched/sched_process_fork`: Tracepoint called after a call for `fork (2)`, `vfork (2)` and `clone (2)`.

78 -- `sched/sched_process_exec`: Tracepoint called after a exec-family syscall.

79 -- `kprobe/kernel_clone`: This is the main [`fork()`](https://elixir.bootlin.com/linux/v5.10/source/kernel/fork.c#L2415)

80 - routine since kernel `5.10.0` was released.

81 -- `kprobe/_do_fork`: Like `kernel_clone`, but this was the main function between kernels `4.2.0` and `5.9.16`

82 -- `kprobe/do_fork`: This was the main function before kernel `4.2.0`.

430 +- `sched/sched_process_fork`: Tracepoint called after a call for `fork (2)`, `vfork (2)` and `clone (2)`.

431 +- `sched/sched_process_exec`: Tracepoint called after a exec-family syscall.

432 +- `kprobe/kernel_clone`: This is the main [`fork()`](https://elixir.bootlin.com/linux/v5.10/source/kernel/fork.c#L2415)

433 + routine since kernel `5.10.0` was released.

434 +- `kprobe/_do_fork`: Like `kernel_clone`, but this was the main function between kernels `4.2.0` and `5.9.16`

435 +- `kprobe/do_fork`: This was the main function before kernel `4.2.0`.

436

437 #### Process Exit

438

@@ -88,9 +441,9 @@ system that the task is finishing its work. The second step is to release the ke

441 function `release_task`. The difference between the two dimensions can help you discover

442 [zombie processes](https://en.wikipedia.org/wiki/Zombie_process). To get the metrics, the collector uses:

443

91 -- `sched/sched_process_exit`: Tracepoint called after a task exits.

92 -- `kprobe/release_task`: This function is called when a process exits, as the kernel still needs to remove the process

93 - descriptor.

444 +- `sched/sched_process_exit`: Tracepoint called after a task exits.

445 +- `kprobe/release_task`: This function is called when a process exits, as the kernel still needs to remove the process

446 + descriptor.

447

448 #### Task error

449

@@ -100,7 +453,7 @@ process and thread creation only.

453 #### Swap

454

455 Inside the swap submenu the eBPF plugin creates the chart `swapcalls`; this chart is displaying when processes are

103 -calling functions [`swap_readpage` and `swap_writepage`](https://hzliu123.github.io/linux-kernel/Page%20Cache%20in%20Linux%202.6.pdf ),

456 +calling functions [`swap_readpage` and `swap_writepage`](https://hzliu123.github.io/linux-kernel/Page%20Cache%20in%20Linux%202.6.pdf),

457 which are functions responsible for doing IO in swap memory. To collect the exact moment that an access to swap happens,

458 the collector attaches `kprobes` for cited functions.

459

@@ -108,91 +461,108 @@ the collector attaches `kprobes` for cited functions.

461

462 The following `tracepoints` are used to measure time usage for soft IRQs:

463

111 -- [`irq/softirq_entry`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_softirq_entry): Called

112 - before softirq handler

113 -- [`irq/softirq_exit`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_softirq_exit): Called when

114 - softirq handler returns.

464 +- [`irq/softirq_entry`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_softirq_entry): Called

465 + before softirq handler

466 +- [`irq/softirq_exit`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_softirq_exit): Called when

467 + softirq handler returns.

468

469 #### Hard IRQ

470

471 The following tracepoints are used to measure the latency of servicing a

472 hardware interrupt request (hard IRQ).

473

121 -- [`irq/irq_handler_entry`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_irq_handler_entry):

122 - Called immediately before the IRQ action handler.

123 -- [`irq/irq_handler_exit`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_irq_handler_exit):

124 - Called immediately after the IRQ action handler returns.

125 -- `irq_vectors`: These are traces from `irq_handler_entry` and

126 - `irq_handler_exit` when an IRQ is handled. The following elements from vector

127 - are triggered:

128 - - `irq_vectors/local_timer_entry`

129 - - `irq_vectors/local_timer_exit`

130 - - `irq_vectors/reschedule_entry`

131 - - `irq_vectors/reschedule_exit`

132 - - `irq_vectors/call_function_entry`

133 - - `irq_vectors/call_function_exit`

134 - - `irq_vectors/call_function_single_entry`

135 - - `irq_vectors/call_function_single_xit`

136 - - `irq_vectors/irq_work_entry`

137 - - `irq_vectors/irq_work_exit`

138 - - `irq_vectors/error_apic_entry`

139 - - `irq_vectors/error_apic_exit`

140 - - `irq_vectors/thermal_apic_entry`

141 - - `irq_vectors/thermal_apic_exit`

142 - - `irq_vectors/threshold_apic_entry`

143 - - `irq_vectors/threshold_apic_exit`

144 - - `irq_vectors/deferred_error_entry`

145 - - `irq_vectors/deferred_error_exit`

146 - - `irq_vectors/spurious_apic_entry`

147 - - `irq_vectors/spurious_apic_exit`

148 - - `irq_vectors/x86_platform_ipi_entry`

149 - - `irq_vectors/x86_platform_ipi_exit`

474 +- [`irq/irq_handler_entry`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_irq_handler_entry):

475 + Called immediately before the IRQ action handler.

476 +- [`irq/irq_handler_exit`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_irq_handler_exit):

477 + Called immediately after the IRQ action handler returns.

478 +- `irq_vectors`: These are traces from `irq_handler_entry` and

479 + `irq_handler_exit` when an IRQ is handled. The following elements from vector

480 + are triggered:

481 + - `irq_vectors/local_timer_entry`

482 + - `irq_vectors/local_timer_exit`

483 + - `irq_vectors/reschedule_entry`

484 + - `irq_vectors/reschedule_exit`

485 + - `irq_vectors/call_function_entry`

486 + - `irq_vectors/call_function_exit`

487 + - `irq_vectors/call_function_single_entry`

488 + - `irq_vectors/call_function_single_xit`

489 + - `irq_vectors/irq_work_entry`

490 + - `irq_vectors/irq_work_exit`

491 + - `irq_vectors/error_apic_entry`

492 + - `irq_vectors/error_apic_exit`

493 + - `irq_vectors/thermal_apic_entry`

494 + - `irq_vectors/thermal_apic_exit`

495 + - `irq_vectors/threshold_apic_entry`

496 + - `irq_vectors/threshold_apic_exit`

497 + - `irq_vectors/deferred_error_entry`

498 + - `irq_vectors/deferred_error_exit`

499 + - `irq_vectors/spurious_apic_entry`

500 + - `irq_vectors/spurious_apic_exit`

501 + - `irq_vectors/x86_platform_ipi_entry`

502 + - `irq_vectors/x86_platform_ipi_exit`

503

504 #### IPC shared memory

505

153 -To monitor shared memory system call counts, the following `kprobes` are used:

506 +To monitor shared memory system call counts, Netdata attaches tracing in the following functions:

507

155 -- `shmget`: Runs when [`shmget`](https://man7.org/linux/man-pages/man2/shmget.2.html) is called.

156 -- `shmat`: Runs when [`shmat`](https://man7.org/linux/man-pages/man2/shmat.2.html) is called.

157 -- `shmdt`: Runs when [`shmdt`](https://man7.org/linux/man-pages/man2/shmat.2.html) is called.

158 -- `shmctl`: Runs when [`shmctl`](https://man7.org/linux/man-pages/man2/shmctl.2.html) is called.

508 +- `shmget`: Runs when [`shmget`](https://man7.org/linux/man-pages/man2/shmget.2.html) is called.

509 +- `shmat`: Runs when [`shmat`](https://man7.org/linux/man-pages/man2/shmat.2.html) is called.

510 +- `shmdt`: Runs when [`shmdt`](https://man7.org/linux/man-pages/man2/shmat.2.html) is called.

511 +- `shmctl`: Runs when [`shmctl`](https://man7.org/linux/man-pages/man2/shmctl.2.html) is called.

512

513 ### Memory

514

515 In the memory submenu the eBPF plugin creates two submenus **page cache** and **synchronization** with the following

516 organization:

517

165 -* Page Cache

166 - * Page cache ratio

167 - * Dirty pages

168 - * Page cache hits

169 - * Page cache misses

170 -* Synchronization

171 - * File sync

172 - * Memory map sync

173 - * File system sync

174 - * File range sync

518 +- Page Cache

519 + - Page cache ratio

520 + - Dirty pages

521 + - Page cache hits

522 + - Page cache misses

523 +- Synchronization

524 + - File sync

525 + - Memory map sync

526 + - File system sync

527 + - File range sync

528

176 -#### Page cache ratio

529 +#### Page cache hits

530

178 -The chart `cachestat_ratio` shows how processes are accessing page cache. In a normal scenario, we expect values around

179 -100%, which means that the majority of the work on the machine is processed in memory. To calculate the ratio, Netdata

180 -attaches `kprobes` for kernel functions:

531 +When the processor needs to read or write a location in main memory, it checks for a corresponding entry in the page cache.

532 + If the entry is there, a page cache hit has occurred and the read is from the cache.

533

182 -- `add_to_page_cache_lru`: Page addition.

183 -- `mark_page_accessed`: Access to cache.

184 -- `account_page_dirtied`: Dirty (modified) pages.

185 -- `mark_buffer_dirty`: Writes to page cache.

534 +A page cache hit is when the page cache is successfully accessed with a read operation. We do not count pages that were

535 +added relatively recently.

536

537 #### Dirty pages

538

539 +A "dirty page" is a page in the page cache that was modified after being created. Since non-dirty pages in the page cache

540 + have identical copies in secondary storage (e.g. hard disk drive or solid-state drive), discarding and reusing their space

541 + is much quicker than paging out application memory, and is often preferred over flushing the dirty pages into secondary storage

542 + and reusing their space.

543 +

544 On `cachestat_dirties` Netdata demonstrates the number of pages that were modified. This chart shows the number of calls

545 to the function `mark_buffer_dirty`.

546

192 -#### Page cache hits

547 +#### Page cache ratio

548

194 -A page cache hit is when the page cache is successfully accessed with a read operation. We do not count pages that were

195 -added relatively recently.

549 +When the processor needs to read or write in a specific memory address, it checks for a corresponding entry in the page cache.

550 +If the processor hits a page cache (`page cache hit`), it reads the entry from the cache. If there is no entry (`page cache miss`),

551 + the kernel allocates a new entry and copies data from the disk. Netdata calculates the percentage of accessed files that are cached on

552 + memory. The ratio is calculated counting the accessed cached pages

553 + (without counting [dirty pages](#dirty-pages) and pages added because of read misses) divided by total access without dirty pages.

554 +

555 +> \_\_**\_\_\_\_**<ins>Number of accessed cached pages</ins>\***\*\_\_\*\***<br/>

556 +> Number of total accessed pages - dirty pages - missed pages

557 +

558 +The chart `cachestat_ratio` shows how processes are accessing page cache. In a normal scenario, we expect values around

559 +100%, which means that the majority of the work on the machine is processed in memory. To calculate the ratio, Netdata

560 +attaches `kprobes` for kernel functions:

561 +

562 +- `add_to_page_cache_lru`: Page addition.

563 +- `mark_page_accessed`: Access to cache.

564 +- `account_page_dirtied`: Dirty (modified) pages.

565 +- `mark_buffer_dirty`: Writes to page cache.

566

567 #### Page cache misses

568

@@ -213,7 +583,7 @@ changes to a file that was mapped into memory using [`mmap(2)`](https://man7.org

583

584 #### File system sync

585

216 -This chart monitors calls demonstrating commits from filesystem caches to disk. Netdata attaches `kprobes` for

586 +This chart monitors calls demonstrating commits from filesystem caches to disk. Netdata attaches `tracing` for

587 [`sync(2)`](https://man7.org/linux/man-pages/man2/sync.2.html), and [`syncfs(2)`](https://man7.org/linux/man-pages/man2/sync.2.html).

588

589 #### File range sync

@@ -240,19 +610,20 @@ By default, MD flush is disabled. To enable it, configure your

610

611 To collect data related to Linux multi-device (MD) flushing, the following kprobe is used:

612

243 -- `kprobe/md_flush_request`: called whenever a request for flushing multi-device data is made.

613 +- `kprobe/md_flush_request`: called whenever a request for flushing multi-device data is made.

614

615 ### Disk

616

617 The eBPF plugin also shows a chart in the Disk section when the `disk` thread is enabled.

618

619 #### Disk Latency

620 +

621 This will create the chart `disk_latency_io` for each disk on the host. The following tracepoints are used:

622

252 -- [`block/block_rq_issue`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_block_rq_issue):

253 - IO request operation to a device drive.

254 -- [`block/block_rq_complete`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_block_rq_complete):

255 - IO operation completed by device.

623 +- [`block/block_rq_issue`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_block_rq_issue):

624 + IO request operation to a device drive.

625 +- [`block/block_rq_complete`](https://www.kernel.org/doc/html/latest/core-api/tracepoint.html#c.trace_block_rq_complete):

626 + IO operation completed by device.

627

628 Disk Latency is the single most important metric to focus on when it comes to storage performance, under most circumstances.

629 For hard drives, an average latency somewhere between 10 to 20 ms can be considered acceptable. For SSD (Solid State Drives),

@@ -265,6 +636,7 @@ This group has charts demonstrating how applications interact with the Linux ker

636 It also brings latency charts for several different filesystems.

637

638 #### Latency Algorithm

639 +

640 We calculate the difference between the calling and return times, spanning disk I/O, file system operations (lock, I/O),

641 run queue latency and all events related to the monitored action.

642

@@ -275,10 +647,10 @@ To measure the latency of executing some actions in an

647 collector needs to attach `kprobes` and `kretprobes` for each of the following

648 functions:

649

278 -- `ext4_file_read_iter`: Function used to measure read latency.

279 -- `ext4_file_write_iter`: Function used to measure write latency.

280 -- `ext4_file_open`: Function used to measure open latency.

281 -- `ext4_sync_file`: Function used to measure sync latency.

650 +- `ext4_file_read_iter`: Function used to measure read latency.

651 +- `ext4_file_write_iter`: Function used to measure write latency.

652 +- `ext4_file_open`: Function used to measure open latency.

653 +- `ext4_sync_file`: Function used to measure sync latency.

654

655 #### ZFS

656

@@ -286,10 +658,10 @@ To measure the latency of executing some actions in a zfs filesystem, the

658 collector needs to attach `kprobes` and `kretprobes` for each of the following

659 functions:

660

289 -- `zpl_iter_read`: Function used to measure read latency.

290 -- `zpl_iter_write`: Function used to measure write latency.

291 -- `zpl_open`: Function used to measure open latency.

292 -- `zpl_fsync`: Function used to measure sync latency.

661 +- `zpl_iter_read`: Function used to measure read latency.

662 +- `zpl_iter_write`: Function used to measure write latency.

663 +- `zpl_open`: Function used to measure open latency.

664 +- `zpl_fsync`: Function used to measure sync latency.

665

666 #### XFS

667

@@ -298,10 +670,10 @@ To measure the latency of executing some actions in an

670 collector needs to attach `kprobes` and `kretprobes` for each of the following

671 functions:

672

301 -- `xfs_file_read_iter`: Function used to measure read latency.

302 -- `xfs_file_write_iter`: Function used to measure write latency.

303 -- `xfs_file_open`: Function used to measure open latency.

304 -- `xfs_file_fsync`: Function used to measure sync latency.

673 +- `xfs_file_read_iter`: Function used to measure read latency.

674 +- `xfs_file_write_iter`: Function used to measure write latency.

675 +- `xfs_file_open`: Function used to measure open latency.

676 +- `xfs_file_fsync`: Function used to measure sync latency.

677

678 #### NFS

679

@@ -310,11 +682,11 @@ To measure the latency of executing some actions in an

682 collector needs to attach `kprobes` and `kretprobes` for each of the following

683 functions:

684

313 -- `nfs_file_read`: Function used to measure read latency.

314 -- `nfs_file_write`: Function used to measure write latency.

315 -- `nfs_file_open`: Functions used to measure open latency.

316 -- `nfs4_file_open`: Functions used to measure open latency for NFS v4.

317 -- `nfs_getattr`: Function used to measure sync latency.

685 +- `nfs_file_read`: Function used to measure read latency.

686 +- `nfs_file_write`: Function used to measure write latency.

687 +- `nfs_file_open`: Functions used to measure open latency.

688 +- `nfs4_file_open`: Functions used to measure open latency for NFS v4.

689 +- `nfs_getattr`: Function used to measure sync latency.

690

691 #### btrfs

692

@@ -322,26 +694,26 @@ To measure the latency of executing some actions in a [btrfs](https://elixir.boo

694 filesystem, the collector needs to attach `kprobes` and `kretprobes` for each of the following functions:

695

696 > Note: We are listing two functions used to measure `read` latency, but we use either `btrfs_file_read_iter` or

325 -`generic_file_read_iter`, depending on kernel version.

697 +> `generic_file_read_iter`, depending on kernel version.

698

327 -- `btrfs_file_read_iter`: Function used to measure read latency since kernel `5.10.0`.

328 -- `generic_file_read_iter`: Like `btrfs_file_read_iter`, but this function was used before kernel `5.10.0`.

329 -- `btrfs_file_write_iter`: Function used to write data.

330 -- `btrfs_file_open`: Function used to open files.

331 -- `btrfs_sync_file`: Function used to synchronize data to filesystem.

699 +- `btrfs_file_read_iter`: Function used to measure read latency since kernel `5.10.0`.

700 +- `generic_file_read_iter`: Like `btrfs_file_read_iter`, but this function was used before kernel `5.10.0`.

701 +- `btrfs_file_write_iter`: Function used to write data.

702 +- `btrfs_file_open`: Function used to open files.

703 +- `btrfs_sync_file`: Function used to synchronize data to filesystem.

704

705 #### File descriptor

706

707 To give metrics related to `open` and `close` events, instead of attaching kprobes for each syscall used to do these

708 events, the collector attaches `kprobes` for the common function used for syscalls:

709

338 -- [`do_sys_open`](https://0xax.gitbooks.io/linux-insides/content/SysCall/linux-syscall-5.html ): Internal function used to

339 - open files.

340 -- [`do_sys_openat2`](https://elixir.bootlin.com/linux/v5.6/source/fs/open.c#L1162):

341 - Function called from `do_sys_open` since version `5.6.0`.

342 -- [`close_fd`](https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg2271761.html): Function used to close file

343 - descriptor since kernel `5.11.0`.

344 -- `__close_fd`: Function used to close files before version `5.11.0`.

710 +- [`do_sys_open`](https://0xax.gitbooks.io/linux-insides/content/SysCall/linux-syscall-5.html): Internal function used to

711 + open files.

712 +- [`do_sys_openat2`](https://elixir.bootlin.com/linux/v5.6/source/fs/open.c#L1162):

713 + Function called from `do_sys_open` since version `5.6.0`.

714 +- [`close_fd`](https://www.mail-archive.com/linux-kernel@vger.kernel.org/msg2271761.html): Function used to close file

715 + descriptor since kernel `5.11.0`.

716 +- `__close_fd`: Function used to close files before version `5.11.0`.

717

718 #### File error

719

@@ -361,22 +733,22 @@ To measure the latency and total quantity of executing some VFS-level

733 functions, ebpf.plugin needs to attach kprobes and kretprobes for each of the

734 following functions:

735

364 -- `vfs_write`: Function used monitoring the number of successful & failed

365 - filesystem write calls, as well as the total number of written bytes.

366 -- `vfs_writev`: Same function as `vfs_write` but for vector writes (i.e. a

367 - single write operation using a group of buffers rather than 1).

368 -- `vfs_read`: Function used for monitoring the number of successful & failed

369 - filesystem read calls, as well as the total number of read bytes.

370 -- `vfs_readv` Same function as `vfs_read` but for vector reads (i.e. a single

371 - read operation using a group of buffers rather than 1).

372 -- `vfs_unlink`: Function used for monitoring the number of successful & failed

373 - filesystem unlink calls.

374 -- `vfs_fsync`: Function used for monitoring the number of successful & failed

375 - filesystem fsync calls.

376 -- `vfs_open`: Function used for monitoring the number of successful & failed

377 - filesystem open calls.

378 -- `vfs_create`: Function used for monitoring the number of successful & failed

379 - filesystem create calls.

736 +- `vfs_write`: Function used monitoring the number of successful & failed

737 + filesystem write calls, as well as the total number of written bytes.

738 +- `vfs_writev`: Same function as `vfs_write` but for vector writes (i.e. a

739 + single write operation using a group of buffers rather than 1).

740 +- `vfs_read`: Function used for monitoring the number of successful & failed

741 + filesystem read calls, as well as the total number of read bytes.

742 +- `vfs_readv` Same function as `vfs_read` but for vector reads (i.e. a single

743 + read operation using a group of buffers rather than 1).

744 +- `vfs_unlink`: Function used for monitoring the number of successful & failed

745 + filesystem unlink calls.

746 +- `vfs_fsync`: Function used for monitoring the number of successful & failed

747 + filesystem fsync calls.

748 +- `vfs_open`: Function used for monitoring the number of successful & failed

749 + filesystem open calls.

750 +- `vfs_create`: Function used for monitoring the number of successful & failed

751 + filesystem create calls.

752

753 ##### VFS Deleted objects

754

@@ -416,35 +788,49 @@ Metrics for directory cache are collected using kprobe for `lookup_fast`, becaus

788 times this function is accessed. On the other hand, for `d_lookup` we are not only interested in the number of times it

789 is accessed, but also in possible errors, so we need to attach a `kretprobe`. For this reason, the following is used:

790

419 -- [`lookup_fast`](https://lwn.net/Articles/649115/): Called to look at data inside the directory cache.

420 -- [`d_lookup`](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/fs/dcache.c?id=052b398a43a7de8c68c13e7fa05d6b3d16ce6801#n2223):

421 - Called when the desired file is not inside the directory cache.

791 +- [`lookup_fast`](https://lwn.net/Articles/649115/): Called to look at data inside the directory cache.

792 +- [`d_lookup`](https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/fs/dcache.c?id=052b398a43a7de8c68c13e7fa05d6b3d16ce6801#n2223):

793 + Called when the desired file is not inside the directory cache.

794

795 ##### Directory Cache Interpretation

796 +

797 When directory cache is showing 100% that means that every accessed file was present in the directory cache.

798 If files are not present in the directory cache, they are either not present in the file system or the files were not

799 accessed before.

800

801 ### Mount Points

802

430 -The following `kprobes` are used to collect `mount` & `unmount` call counts:

803 +The following `tracing` are used to collect `mount` & `unmount` call counts:

804

432 -- [`mount`](https://man7.org/linux/man-pages/man2/mount.2.html): mount filesystem on host.

433 -- [`umount`](https://man7.org/linux/man-pages/man2/umount.2.html): umount filesystem on host.

805 +- [`mount`](https://man7.org/linux/man-pages/man2/mount.2.html): mount filesystem on host.

806 +- [`umount`](https://man7.org/linux/man-pages/man2/umount.2.html): umount filesystem on host.

807

808 ### Networking Stack

809

437 -Netdata monitors socket bandwidth attaching `kprobes` for internal functions.

810 +Netdata monitors socket bandwidth attaching `tracing` for internal functions.

811 +

812 +#### TCP outbound connections

813 +

814 +This chart demonstrates calls to `tcp_v4_connection` and `tcp_v6_connection` that start connections for IPV4 and IPV6, respectively.

815

439 -#### TCP functions

816 +#### TCP inbound connections

817 +

818 +This chart demonstrates TCP and UDP connections that the host receives.

819 +To collect this information, netdata attaches a tracing to `inet_csk_accept`.

820 +

821 +#### TCP bandwidth functions

822

823 This chart demonstrates calls to functions `tcp_sendmsg`, `tcp_cleanup_rbuf`, and `tcp_close`; these functions are used

824 to send & receive data and to close connections when `TCP` protocol is used.

825

826 #### TCP bandwidth

827

446 -Like the previous chart, this one also monitors `tcp_sendmsg` and `tcp_cleanup_rbuf`, but instead of showing the number

447 -of calls, it demonstrates the number of bytes sent and received.

828 +This chart demonstrates calls to functions:

829 +

830 +- `tcp_sendmsg`: Function responsible to send data for a specified destination.

831 +- `tcp_cleanup_rbuf`: We use this function instead of `tcp_recvmsg`, because the last one misses `tcp_read_sock` traffic

832 + and we would also need to add more `tracing` to get the socket and package size.

833 +- `tcp_close`: Function responsible to close connection.

834

835 #### TCP retransmit

836

@@ -467,359 +853,11 @@ calls, it monitors the number of bytes sent and received.

853

854 These are tracepoints related to [OOM](https://en.wikipedia.org/wiki/Out_of_memory) killing processes.

855

470 -- `oom/mark_victim`: Monitors when an oomkill event happens.

471 -

472 -## Configuration

473 -

474 -Enable or disable the entire eBPF collector by editing `netdata.conf`.

475 -

476 -```bash

477 -cd /etc/netdata/ # Replace with your Netdata configuration directory, if not /etc/netdata/

478 -./edit-config netdata.conf

479 -```

480 -

481 -To enable the collector, scroll down to the `[plugins]` section ensure the relevant line references `ebpf` (not

482 -`ebpf_process`), is uncommented, and is set to `yes`.

483 -

484 -```conf

485 -[plugins]

486 - ebpf = yes

487 -```

488 -

489 -You can also configure the eBPF collector's behavior by editing `ebpf.d.conf`.

490 -

491 -```bash

492 -cd /etc/netdata/ # Replace with your Netdata configuration directory, if not /etc/netdata/

493 -./edit-config ebpf.d.conf

494 -```

495 -

496 -### `[global]`

497 -

498 -The `[global]` section defines settings for the whole eBPF collector.

499 -

500 -#### eBPF load mode

501 -

502 -The collector has two different eBPF programs. These programs monitor the same functions inside the kernel, but they

503 -monitor, process, and display different kinds of information.

504 -

505 -By default, this plugin uses the `entry` mode. Changing this mode can create significant overhead on your operating

506 -system, but also offer valuable information if you are developing or debugging software. The `ebpf load mode` option

507 -accepts the following values:

508 -

509 -- `entry`: This is the default mode. In this mode, the eBPF collector only monitors calls for the functions described in

510 - the sections above, and does not show charts related to errors.

511 -- `return`: In the `return` mode, the eBPF collector monitors the same kernel functions as `entry`, but also creates new

512 - charts for the return of these functions, such as errors. Monitoring function returns can help in debugging software,

513 - such as failing to close file descriptors or creating zombie processes.

514 -- `update every`: Number of seconds used for eBPF to send data for Netdata.

515 -- `pid table size`: Defines the maximum number of PIDs stored inside the application hash table.

516 -

517 -#### Integration with `apps.plugin`

518 -

519 -The eBPF collector also creates charts for each running application through an integration with the

520 -[`apps.plugin`](/collectors/apps.plugin/README.md). This integration helps you understand how specific applications

521 -interact with the Linux kernel.

522 -

523 -If you want to _disable_ the integration with `apps.plugin` along with the above charts, change the setting `apps` to

524 -`no`.

525 -

526 -```conf

527 -[global]

528 - apps = yes

529 -```

530 -

531 -When the integration is enabled, eBPF collector allocates memory for each process running. The total allocated memory

532 -has direct relationship with the kernel version. When the eBPF plugin is running on kernels newer than `4.15`, it uses

533 -per-cpu maps to speed up the update of hash tables. This also implies storing data for the same PID for each processor

534 -it runs.

535 -

536 -#### Integration with `cgroups.plugin`

537 -

538 -The eBPF collector also creates charts for each cgroup through an integration with the

539 -[`cgroups.plugin`](/collectors/cgroups.plugin/README.md). This integration helps you understand how a specific cgroup

540 -interacts with the Linux kernel.

541 -

542 -The integration with `cgroups.plugin` is disabled by default to avoid creating overhead on your system. If you want to

543 -_enable_ the integration with `cgroups.plugin`, change the `cgroups` setting to `yes`.

544 -

545 -```conf

546 -[global]

547 - cgroups = yes

548 -```

549 -

550 -If you do not need to monitor specific metrics for your `cgroups`, you can enable `cgroups` inside

551 -`ebpf.d.conf`, and then disable the plugin for a specific `thread` by following the steps in the

552 -[Configuration](#configuration) section.

553 -

554 -#### Integration Dashboard Elements

555 -

556 -When an integration is enabled, your dashboard will also show the following cgroups and apps charts using low-level

557 -Linux metrics:

558 -

559 -> Note: The parenthetical accompanying each bulleted item provides the chart name.

560 -

561 -- mem

562 - - Number of processes killed due out of memory. (`oomkills`)

563 -- process

564 - - Number of processes created with `do_fork`. (`process_create`)

565 - - Number of threads created with `do_fork` or `clone (2)`, depending on your system's kernel

566 - version. (`thread_create`)

567 - - Number of times that a process called `do_exit`. (`task_exit`)

568 - - Number of times that a process called `release_task`. (`task_close`)

569 - - Number of times that an error happened to create thread or process. (`task_error`)

570 -- swap

571 - - Number of calls to `swap_readpage`. (`swap_read_call`)

572 - - Number of calls to `swap_writepage`. (`swap_write_call`)

573 -- network

574 - - Number of bytes sent. (`total_bandwidth_sent`)

575 - - Number of bytes received. (`total_bandwidth_recv`)

576 - - Number of calls to `tcp_sendmsg`. (`bandwidth_tcp_send`)

577 - - Number of calls to `tcp_cleanup_rbuf`. (`bandwidth_tcp_recv`)

578 - - Number of calls to `tcp_retransmit_skb`. (`bandwidth_tcp_retransmit`)

579 - - Number of calls to `udp_sendmsg`. (`bandwidth_udp_send`)

580 - - Number of calls to `udp_recvmsg`. (`bandwidth_udp_recv`)

581 -- file access

582 - - Number of calls to open files. (`file_open`)

583 - - Number of calls to open files that returned errors. (`open_error`)

584 - - Number of files closed. (`file_closed`)

585 - - Number of calls to close files that returned errors. (`file_error_closed`)

586 -- vfs

587 - - Number of calls to `vfs_unlink`. (`file_deleted`)

588 - - Number of calls to `vfs_write`. (`vfs_write_call`)

589 - - Number of calls to write a file that returned errors. (`vfs_write_error`)

590 - - Number of calls to `vfs_read`. (`vfs_read_call`)

591 - - Number of bytes written with `vfs_write`. (`vfs_write_bytes`)

592 - - Number of bytes read with `vfs_read`. (`vfs_read_bytes`)

593 - - Number of calls to read a file that returned errors. (`vfs_read_error`)

594 - - Number of calls to `vfs_fsync`. (`vfs_fsync`)

595 - - Number of calls to sync file that returned errors. (`vfs_fsync_error`)

596 - - Number of calls to `vfs_open`. (`vfs_open`)

597 - - Number of calls to open file that returned errors. (`vfs_open_error`)

598 - - Number of calls to `vfs_create`. (`vfs_create`)

599 - - Number of calls to open file that returned errors. (`vfs_create_error`)

600 -- page cache

601 - - Ratio of pages accessed. (`cachestat_ratio`)

602 - - Number of modified pages ("dirty"). (`cachestat_dirties`)

603 - - Number of accessed pages. (`cachestat_hits`)

604 - - Number of pages brought from disk. (`cachestat_misses`)

605 -- directory cache

606 - - Ratio of files available in directory cache. (`dc_hit_ratio`)

607 - - Number of files accessed. (`dc_reference`)

608 - - Number of files accessed that were not in cache. (`dc_not_cache`)

609 - - Number of files not found. (`dc_not_found`)

610 -- ipc shm

611 - - Number of calls to `shm_get`. (`shmget_call`)

612 - - Number of calls to `shm_at`. (`shmat_call`)

613 - - Number of calls to `shm_dt`. (`shmdt_call`)

614 - - Number of calls to `shm_ctl`. (`shmctl_call`)

615 -

616 -### `[ebpf programs]`

617 -

618 -The eBPF collector enables and runs the following eBPF programs by default:

619 -

620 -- `fd` : This eBPF program creates charts that show information about calls to open files.

621 -- `mount`: This eBPF program creates charts that show calls to syscalls mount(2) and umount(2).

622 -- `shm`: This eBPF program creates charts that show calls to syscalls shmget(2), shmat(2), shmdt(2) and shmctl(2).

623 -- `sync`: Monitor calls to syscalls sync(2), fsync(2), fdatasync(2), syncfs(2), msync(2), and sync_file_range(2).

624 -- `network viewer`: This eBPF program creates charts with information about `TCP` and `UDP` functions, including the

625 - bandwidth consumed by each.

626 -- `vfs`: This eBPF program creates charts that show information about VFS (Virtual File System) functions.

627 -- `process`: This eBPF program creates charts that show information about process life. When in `return` mode, it also

628 - creates charts showing errors when these operations are executed.

629 -- `hardirq`: This eBPF program creates charts that show information about time spent servicing individual hardware

630 - interrupt requests (hard IRQs).

631 -- `softirq`: This eBPF program creates charts that show information about time spent servicing individual software

632 - interrupt requests (soft IRQs).

633 -- `oomkill`: This eBPF program creates a chart that shows OOM kills for all applications recognized via

634 - the `apps.plugin` integration. Note that this program will show application charts regardless of whether apps

635 - integration is turned on or off.

636 -

637 -You can also enable the following eBPF programs:

638 -

639 -- `cachestat`: Netdata's eBPF data collector creates charts about the memory page cache. When the integration with

640 - [`apps.plugin`](/collectors/apps.plugin/README.md) is enabled, this collector creates charts for the whole host _and_

641 - for each application.

642 -- `dcstat` : This eBPF program creates charts that show information about file access using directory cache. It appends

643 - `kprobes` for `lookup_fast()` and `d_lookup()` to identify if files are inside directory cache, outside and files are

644 - not found.

645 -- `disk` : This eBPF program creates charts that show information about disk latency independent of filesystem.

646 -- `filesystem` : This eBPF program creates charts that show information about some filesystem latency.

647 -- `swap` : This eBPF program creates charts that show information about swap access.

648 -- `mdflush`: This eBPF program creates charts that show information about

649 - multi-device software flushes.

650 -

651 -## Thread configuration

652 -

653 -You can configure each thread of the eBPF data collector by editing either the `cachestat.conf`, `process.conf`,

654 -or `network.conf` files. Use [`edit-config`](/docs/configure/nodes.md) from your Netdata config directory:

655 -

656 -```bash

657 -cd /etc/netdata/ # Replace with your Netdata configuration directory, if not /etc/netdata/

658 -./edit-config ebpf.d/process.conf

659 -```

660 -

661 -### Configuration files

662 -

663 -The following configuration files are available:

664 -

665 -- `cachestat.conf`: Configuration for the `cachestat` thread.

666 -- `dcstat.conf`: Configuration for the `dcstat` thread.

667 -- `disk.conf`: Configuration for the `disk` thread.

668 -- `fd.conf`: Configuration for the `file descriptor` thread.

669 -- `filesystem.conf`: Configuration for the `filesystem` thread.

670 -- `hardirq.conf`: Configuration for the `hardirq` thread.

671 -- `process.conf`: Configuration for the `process` thread.

672 -- `network.conf`: Configuration for the `network viewer` thread. This config file overwrites the global options and also

673 - lets you specify which network the eBPF collector monitors.

674 -- `softirq.conf`: Configuration for the `softirq` thread.

675 -- `sync.conf`: Configuration for the `sync` thread.

676 -- `vfs.conf`: Configuration for the `vfs` thread.

677 -

678 -### Network configuration

679 -

680 -The network configuration has specific options to configure which network(s) the eBPF collector monitors. These options

681 -are divided in the following sections:

682 -

683 -#### `[network connections]`

684 -

685 -You can configure the information shown on `outbound` and `inbound` charts with the settings in this section.

686 -

687 -```conf

688 -[network connections]

689 - maximum dimensions = 500

690 - resolve hostname ips = no

691 - ports = 1-1024 !145 !domain

692 - hostnames = !example.com

693 - ips = !127.0.0.1/8 10.0.0.0/8 172.16.0.0/12 192.168.0.0/16 fc00::/7

694 -```

695 -

696 -When you define a `ports` setting, Netdata will collect network metrics for that specific port. For example, if you

697 -write `ports = 19999`, Netdata will collect only connections for itself. The `hostnames` setting accepts

698 -[simple patterns](/libnetdata/simple_pattern/README.md). The `ports`, and `ips` settings accept negation (`!`) to deny

699 -specific values or asterisk alone to define all values.

700 -

701 -In the above example, Netdata will collect metrics for all ports between 1 and 443, with the exception of 53 (domain)

702 -and 145.

703 -

704 -The following options are available:

705 -

706 -- `ports`: Define the destination ports for Netdata to monitor.

707 -- `hostnames`: The list of hostnames that can be resolved to an IP address.

708 -- `ips`: The IP or range of IPs that you want to monitor. You can use IPv4 or IPv6 addresses, use dashes to define a

709 - range of IPs, or use CIDR values. The default behavior is to only collect data for private IP addresses, but this can

710 - be changed with the `ips` setting.

711 -

712 -By default, Netdata displays up to 500 dimensions on network connection charts. If there are more possible dimensions,

713 -they will be bundled into the `other` dimension. You can increase the number of shown dimensions by changing

714 -the `maximum dimensions` setting.

856 +- `oom/mark_victim`: Monitors when an oomkill event happens.

857

716 -The dimensions for the traffic charts are created using the destination IPs of the sockets by default. This can be

717 -changed setting `resolve hostname ips = yes` and restarting Netdata, after this Netdata will create dimensions using

718 -the `hostnames` every time that is possible to resolve IPs to their hostnames.

719 -

720 -#### `[service name]`

721 -

722 -Netdata uses the list of services in `/etc/services` to plot network connection charts. If this file does not contain

723 -the name for a particular service you use in your infrastructure, you will need to add it to the `[service name]`

724 -section.

858 +## Known issues

859

726 -For example, Netdata's default port (`19999`) is not listed in `/etc/services`. To associate that port with the Netdata

727 -service in network connection charts, and thus see the name of the service instead of its port, define it:

728 -

729 -```conf

730 -[service name]

731 - 19999 = Netdata

732 -```

733 -

734 -### Sync configuration

735 -

736 -The sync configuration has specific options to disable monitoring for syscalls, as default option all syscalls are

737 -monitored.

738 -

739 -```conf

740 -[syscalls]

741 - sync = yes

742 - msync = yes

743 - fsync = yes

744 - fdatasync = yes

745 - syncfs = yes

746 - sync_file_range = yes

747 -```

748 -

749 -### Filesystem configuration

750 -

751 -The filesystem configuration has specific options to disable monitoring for filesystems, by default all filesystems are

752 -monitored.

753 -

754 -```conf

755 -[filesystem]

756 - btrfsdist = yes

757 - ext4dist = yes

758 - nfsdist = yes

759 - xfsdist = yes

760 - zfsdist = yes

761 -```

762 -

763 -The ebpf program `nfsdist` monitors only `nfs` mount points.

764 -

765 -## Troubleshooting

766 -

767 -If the eBPF collector does not work, you can troubleshoot it by running the `ebpf.plugin` command and investigating its

768 -output.

769 -

770 -```bash

771 -cd /usr/libexec/netdata/plugins.d/

772 -sudo su -s /bin/bash netdata

773 -./ebpf.plugin

774 -```

775 -

776 -You can also use `grep` to search the Agent's `error.log` for messages related to eBPF monitoring.

777 -

778 -```bash

779 -grep -i ebpf /var/log/netdata/error.log

780 -```

781 -

782 -### Confirm kernel compatibility

783 -

784 -The eBPF collector only works on Linux systems and with specific Linux kernels. We support all kernels more recent than

785 -`4.11.0`, and all kernels on CentOS 7.6 or later.

786 -

787 -You can run our helper script to determine whether your system can support eBPF monitoring.

788 -

789 -```bash

790 -curl -sSL https://raw.githubusercontent.com/netdata/kernel-collector/master/tools/check-kernel-config.sh | sudo bash

791 -```

792 -

793 -If this script returns no output, your system is ready to compile and run the eBPF collector.

794 -

795 -If you see a warning about a missing kernel

796 -configuration (`KPROBES KPROBES_ON_FTRACE HAVE_KPROBES BPF BPF_SYSCALL BPF_JIT`), you will need to recompile your kernel

797 -to support this configuration. The process of recompiling Linux kernels varies based on your distribution and version.

798 -Read the documentation for your system's distribution to learn more about the specific workflow for recompiling the

799 -kernel, ensuring that you set all the necessary

800 -

801 -- [Ubuntu](https://wiki.ubuntu.com/Kernel/BuildYourOwnKernel)

802 -- [Debian](https://kernel-team.pages.debian.net/kernel-handbook/ch-common-tasks.html#s-common-official)

803 -- [Fedora](https://fedoraproject.org/wiki/Building_a_custom_kernel)

804 -- [CentOS](https://wiki.centos.org/HowTos/Custom_Kernel)

805 -- [Arch Linux](https://wiki.archlinux.org/index.php/Kernel/Traditional_compilation)

806 -- [Slackware](https://docs.slackware.com/howtos:slackware_admin:kernelbuilding)

807 -

808 -### Mount `debugfs` and `tracefs`

809 -

810 -The eBPF collector also requires both the `tracefs` and `debugfs` filesystems. Try mounting the `tracefs` and `debugfs`

811 -filesystems using the commands below:

812 -

813 -```bash

814 -sudo mount -t debugfs nodev /sys/kernel/debug

815 -sudo mount -t tracefs nodev /sys/kernel/tracing

816 -```

817 -

818 -If they are already mounted, you will see an error. You can also configure your system's `/etc/fstab` configuration to

819 -mount these filesystems on startup. More information can be found in

820 -the [ftrace documentation](https://www.kernel.org/doc/Documentation/trace/ftrace.txt).

821 -

822 -## Performance

860 +### Performance opimization

861

862 eBPF monitoring is complex and produces a large volume of metrics. We've discovered scenarios where the eBPF plugin

863 significantly increases kernel memory usage by several hundred MB.

@@ -832,7 +870,7 @@ usage (see the `system.ram` chart) has dropped significantly.

870 Beginning with `v1.31`, kernel memory usage is configurable via the [`pid table size` setting](#ebpf-load-mode)

871 in `ebpf.conf`.

872

835 -## SELinux

873 +### SELinux

874

875 When [SELinux](https://www.redhat.com/en/topics/linux/what-is-selinux) is enabled, it may prevent `ebpf.plugin` from

876 starting correctly. Check the Agent's `error.log` file for errors like the ones below:

@@ -852,7 +890,7 @@ ess.pl" exe="/usr/libexec/netdata/plugins.d/ebpf.plugin" subj=system_u:system_r:

890

891 If you see similar errors, you will have to adjust SELinux's policies to enable the eBPF collector.

892

855 -### Creation of bpf policies

893 +#### Creation of bpf policies

894

895 To enable `ebpf.plugin` to run on a distribution with SELinux enabled, it will be necessary to take the following

896 actions.

@@ -898,14 +936,14 @@ Finally, you can load the new policy and start the Netdata agent again:

936 # systemctl start netdata

937 ```

938

901 -## Lockdown

939 +### Linux kernel lockdown

940

941 Beginning with [version 5.4](https://www.zdnet.com/article/linux-to-get-kernel-lockdown-feature/), the Linux kernel has

942 a feature called "lockdown," which may affect `ebpf.plugin` depending how the kernel was compiled. The following table

943 shows how the lockdown module impacts `ebpf.plugin` based on the selected options:

944

908 -|:------------------------- |:--------------------------------- |:--------------------- |:-------------------------------- |

946 +| :------------------------ | :-------------------------------- | :-------------------- | :------------------------------- |

947 | YES | NO | NO | YES |

948 | YES | Yes | None | YES |

949 | YES | Yes | Integrity | YES |

@@ -913,5 +951,3 @@ shows how the lockdown module impacts `ebpf.plugin` based on the selected option

951

952 If you or your distribution compiled the kernel with the last combination, your system cannot load shared libraries

953 required to run `ebpf.plugin`.

916 -

917 -

Update ebpf doc (#12503)