| 1 | lttng-relayd(8) |
| 2 | =============== |
| 3 | :revdate: 5 June 2018 |
| 4 | |
| 5 | |
| 6 | NAME |
| 7 | ---- |
| 8 | lttng-relayd - LTTng 2 relay daemon |
| 9 | |
| 10 | |
| 11 | SYNOPSIS |
| 12 | -------- |
| 13 | [verse] |
| 14 | *lttng-relayd* [option:--background | option:--daemonize] |
| 15 | [option:--control-port='URL'] [option:--data-port='URL'] [option:--fd-pool-size='COUNT'] |
| 16 | [option:--live-port='URL'] [option:--output='PATH'] |
| 17 | [option:-v | option:-vv | option:-vvv] [option:--working-directory='PATH'] |
| 18 | [option:--group-output-by-session | option:--group-output-by-host] |
| 19 | |
| 20 | DESCRIPTION |
| 21 | ----------- |
| 22 | The https://lttng.org/[_Linux Trace Toolkit: next generation_] is an open |
| 23 | source software package used for correlated tracing of the Linux kernel, |
| 24 | user applications, and user libraries. |
| 25 | |
| 26 | LTTng consists of Linux kernel modules (for Linux kernel tracing) and |
| 27 | dynamically loaded libraries (for user application and library tracing). |
| 28 | |
| 29 | The _LTTng relay daemon_ is responsible for receiving trace data from |
| 30 | possibly remote LTTng session/consumer daemons and for writing it to |
| 31 | the local file system. The relay daemon also accepts _LTTng live_ |
| 32 | connections from compatible viewers; this is the official approach to |
| 33 | viewing LTTng events as they are emitted. |
| 34 | |
| 35 | The relay daemon listens by default on all network interfaces to gather |
| 36 | trace data, but only on localhost for LTTng live connections. |
| 37 | |
| 38 | The relay daemon does not require any particular permissions, as long as |
| 39 | it can write to the output directory and listen on the configured ports. |
| 40 | If a user is within a secured network and/or has proper firewall |
| 41 | settings, `lttng-relayd` can listen to LTTng live connections from _all_ |
| 42 | network interfaces by specifying |
| 43 | +--live-port=tcp://0.0.0.0:{default_network_viewer_port}+. |
| 44 | |
| 45 | Once a trace has been streamed completely, the trace can be processed by |
| 46 | any tool that can process an LTTng trace located on the local |
| 47 | file system. |
| 48 | |
| 49 | |
| 50 | [[output-directory]] |
| 51 | Output directory |
| 52 | ~~~~~~~~~~~~~~~~ |
| 53 | By default (option:--group-output-by-host), the relay daemon writes the traces |
| 54 | to: |
| 55 | |
| 56 | [verse] |
| 57 | $LTTNG_HOME/lttng-traces/'HOSTNAME'/'SESSION'/'DOMAIN' |
| 58 | |
| 59 | with: |
| 60 | |
| 61 | 'HOSTNAME':: |
| 62 | Remote hostname. |
| 63 | |
| 64 | 'SESSION':: |
| 65 | Full session name. |
| 66 | |
| 67 | 'DOMAIN':: |
| 68 | Tracing domain. |
| 69 | |
| 70 | You can group the output per session instead of per hostname with the |
| 71 | option:--group-output-by-session option. This result in the following: |
| 72 | |
| 73 | [verse] |
| 74 | |
| 75 | $LTTNG_HOME/lttng-traces/'SESSION'/'HOST'/'DOMAIN' |
| 76 | |
| 77 | You can override the default output directory prefix |
| 78 | (`$LTTNG_HOME/lttng-traces`) with the option:--output option. The other |
| 79 | parts depend on the remote configuration. |
| 80 | |
| 81 | Grouping by session is not supported for peers with version inferior to 2.4. |
| 82 | In such cases, the grouping of output for that particular peer will be per host |
| 83 | (default). |
| 84 | |
| 85 | A best effort backward compatibility is done for peers version greater or equal to |
| 86 | 2.4 and smaller than 2.11. |
| 87 | |
| 88 | [[url-format]] |
| 89 | URL format |
| 90 | ~~~~~~~~~~ |
| 91 | The option:--control-port, option:--data-port, and option:--live-port |
| 92 | options specify URLs. |
| 93 | |
| 94 | The format of those URLs is: |
| 95 | |
| 96 | [verse] |
| 97 | tcp://('HOST' | 'IPADDR'):__PORT__ |
| 98 | |
| 99 | with: |
| 100 | |
| 101 | ('HOST' | 'IPADDR'):: |
| 102 | Binding hostname or IP address (IPv6 address *must* be enclosed in |
| 103 | brackets (`[` and `]`); see |
| 104 | https://www.ietf.org/rfc/rfc2732.txt[RFC 2732]). |
| 105 | |
| 106 | 'PORT':: |
| 107 | TCP port. |
| 108 | |
| 109 | |
| 110 | OPTIONS |
| 111 | ------- |
| 112 | Daemon |
| 113 | ~~~~~~ |
| 114 | option:-b, option:--background:: |
| 115 | Start as Unix daemon, but keep file descriptors (console) open. |
| 116 | Use the option:--daemonize option instead to close the file |
| 117 | descriptors. |
| 118 | |
| 119 | option:-d, option:--daemonize:: |
| 120 | Start as Unix daemon, and close file descriptors (console). Use the |
| 121 | option:--background option instead to keep the file descriptors |
| 122 | open. |
| 123 | |
| 124 | option:--fd-pool-size='SIZE':: |
| 125 | Set the size of the file descriptor pool. This effectively sets a |
| 126 | limit on the number of file descriptors that may be kept open |
| 127 | simultaneously by the daemon (default: the soft `RLIMIT_NOFILE` resource |
| 128 | limit of the process). |
| 129 | |
| 130 | option:-g 'GROUP', option:--group='GROUP':: |
| 131 | Use 'GROUP' as Unix tracing group (default: `tracing`). |
| 132 | |
| 133 | option:-o 'PATH', option:--output='PATH':: |
| 134 | Set base directory of written trace data to 'PATH'. |
| 135 | + |
| 136 | See the <<output-directory,Output directory>> section above for more |
| 137 | information. |
| 138 | |
| 139 | option:-p, option:--group-output-by-host:: |
| 140 | Organize the output directory by host. Default behavior. |
| 141 | + |
| 142 | See the <<output-directory,Output directory>> section above for more |
| 143 | information. |
| 144 | |
| 145 | option:-s, option:--group-output-by-session:: |
| 146 | Organize the output directory by session. |
| 147 | + |
| 148 | See the <<output-directory,Output directory>> section above for more |
| 149 | information. |
| 150 | |
| 151 | option:-w 'PATH', option:--working-directory='PATH':: |
| 152 | Set the working directory of the processes this relay daemon creates. |
| 153 | |
| 154 | option:-v, option:--verbose:: |
| 155 | Increase verbosity. |
| 156 | + |
| 157 | Three levels of verbosity are available, which are triggered by |
| 158 | appending additional `v` letters to the option |
| 159 | (that is, `-vv` and `-vvv`). |
| 160 | |
| 161 | |
| 162 | Ports |
| 163 | ~~~~~ |
| 164 | See the <<url-format,URL format>> section above for more information |
| 165 | about the syntax of the following options' 'URL' argument. |
| 166 | |
| 167 | option:-C 'URL', option:--control-port='URL':: |
| 168 | Listen to control data on URL 'URL' (default: |
| 169 | +tcp://{default_network_control_bind_address}:{default_network_control_port}+). |
| 170 | |
| 171 | option:-D 'URL', option:--data-port='URL':: |
| 172 | Listen to trace data on URL 'URL' (default: |
| 173 | +tcp://{default_network_data_bind_address}:{default_network_data_port}+). |
| 174 | |
| 175 | option:-L 'URL', option:--live-port='URL':: |
| 176 | Listen to LTTng live connections on URL 'URL' |
| 177 | (default: |
| 178 | +tcp://{default_network_viewer_bind_address}:{default_network_viewer_port}+). |
| 179 | |
| 180 | |
| 181 | Program information |
| 182 | ~~~~~~~~~~~~~~~~~~~ |
| 183 | option:-h, option:--help:: |
| 184 | Show help. |
| 185 | |
| 186 | option:-V, option:--version:: |
| 187 | Show version. |
| 188 | |
| 189 | |
| 190 | ENVIRONMENT VARIABLES |
| 191 | --------------------- |
| 192 | `LTTNG_ABORT_ON_ERROR`:: |
| 193 | Set to 1 to abort the process after the first error is encountered. |
| 194 | |
| 195 | `LTTNG_NETWORK_SOCKET_TIMEOUT`:: |
| 196 | Socket connection, receive and send timeout (milliseconds). A value |
| 197 | of 0 or -1 uses the timeout of the operating system (default). |
| 198 | |
| 199 | `LTTNG_RELAYD_DISALLOW_CLEAR`:: |
| 200 | Set to 1 to disallow the use of lttng-clear(1) on sessions collected by |
| 201 | this relay daemon. |
| 202 | |
| 203 | `LTTNG_RELAYD_HEALTH`:: |
| 204 | Path to relay daemon health's socket. |
| 205 | |
| 206 | `LTTNG_RELAYD_TCP_KEEP_ALIVE`:: |
| 207 | Set to 1 to enable TCP keep-alive. |
| 208 | + |
| 209 | The TCP keep-alive mechanism allows the detection of dead peers |
| 210 | (man:lttng-sessiond(8)) in cases of unclean termination |
| 211 | (for example, a hard reset) of a peer. |
| 212 | + |
| 213 | Supported on Linux and Solaris only. The default behaviour of the TCP |
| 214 | keep-alive mechanism is OS-specific. |
| 215 | + |
| 216 | Search for `tcp_keepalive` in man:tcp(7) for more information. |
| 217 | |
| 218 | `LTTNG_RELAYD_TCP_KEEP_ALIVE_ABORT_THRESHOLD`:: |
| 219 | The time threshold in seconds to abort a TCP connection after the keep-alive |
| 220 | probing mechanism has failed. |
| 221 | + |
| 222 | Set to 0 or -1 to use the value chosen by the operating system (default). |
| 223 | + |
| 224 | Supported on Solaris 11 only. |
| 225 | + |
| 226 | Search for `tcp_keepalive_abort_threshold` in man:tcp(7) for more information. |
| 227 | |
| 228 | `LTTNG_RELAYD_TCP_KEEP_ALIVE_IDLE_TIME`:: |
| 229 | Number of seconds a connection needs to be idle before TCP begins |
| 230 | sending out keep-alive probes. |
| 231 | + |
| 232 | Set to 0 or -1 to use the value chosen by the operating system (default). |
| 233 | + |
| 234 | Supported on Linux and Solaris 11 only. |
| 235 | + |
| 236 | On Solaris{nbsp}11, the accepted values are -1, 0, and 10 to 864000. |
| 237 | + |
| 238 | Search for `tcp_keepalive_time` and `tcp_keepalive_interval` |
| 239 | in man:tcp(7) on Solaris 11 for more information. |
| 240 | |
| 241 | `LTTNG_RELAYD_TCP_KEEP_ALIVE_MAX_PROBE_COUNT`:: |
| 242 | Maximum number of TCP keep-alive probes to send before giving up and |
| 243 | killing the connection if no response is obtained from the other end. |
| 244 | + |
| 245 | Set to 0 or -1 to use the value chosen by the operating system (default). |
| 246 | + |
| 247 | Supported on Linux only. |
| 248 | + |
| 249 | Search for `tcp_keepalive_probes` in man:tcp(7) for more information. |
| 250 | |
| 251 | `LTTNG_RELAYD_TCP_KEEP_ALIVE_PROBE_INTERVAL`:: |
| 252 | Number of seconds between TCP keep-alive probes. |
| 253 | + |
| 254 | Set to 0 or -1 to use the value chosen by the operating system (default). |
| 255 | + |
| 256 | Supported on Linux only. |
| 257 | + |
| 258 | Search for `tcp_keepalive_intvl` in man:tcp(7) for more information. |
| 259 | |
| 260 | `LTTNG_RELAYD_WORKING_DIRECTORY`:: |
| 261 | Working directory of the processes this relay daemon creates. |
| 262 | + |
| 263 | The option:--working-directory option overrides this variable. |
| 264 | |
| 265 | |
| 266 | FILES |
| 267 | ----- |
| 268 | `$LTTNG_HOME/.lttng`:: |
| 269 | User LTTng runtime and configuration directory. |
| 270 | |
| 271 | `$LTTNG_HOME/lttng-traces`:: |
| 272 | Default output directory of LTTng traces. This can be overridden |
| 273 | with the option:--output option. |
| 274 | |
| 275 | NOTE: `$LTTNG_HOME` defaults to `$HOME` when not explicitly set. |
| 276 | |
| 277 | |
| 278 | EXIT STATUS |
| 279 | ----------- |
| 280 | *0*:: |
| 281 | Success |
| 282 | |
| 283 | *1*:: |
| 284 | Error |
| 285 | |
| 286 | *3*:: |
| 287 | Fatal error |
| 288 | |
| 289 | |
| 290 | LIMITATIONS |
| 291 | ----------- |
| 292 | As of this version, only the TCP protocol is supported for both control |
| 293 | and data ports. In future versions, TCP will remain the sole available |
| 294 | protocol for control data since those communications are low-volume and |
| 295 | need absolute reliability; trace data could be carried over UDP. |
| 296 | |
| 297 | For an unprivileged user running `lttng-relayd`, the maximum number of |
| 298 | file descriptors per process is usually 1024. This limits the number of |
| 299 | connections and opened trace files. This limit can be configured with |
| 300 | *ulimit*(3). |
| 301 | |
| 302 | |
| 303 | include::common-footer.txt[] |
| 304 | |
| 305 | |
| 306 | SEE ALSO |
| 307 | -------- |
| 308 | man:lttng(1), |
| 309 | man:lttng-sessiond(8), |
| 310 | man:lttng-crash(1), |
| 311 | man:lttng-ust(3), |
| 312 | man:babeltrace(1) |