kernel(7)

📦
version 25.3.2.8 · 3 versions
, in ubuntu noble

✨ kernel(7) — Erlang Application Definition ✨

💖 NAME 💖

kernel — the Kernel application. That's the name, bestie, and she's iconic. 💅

🌸 DESCRIPTION 🌸

Okay so the Kernel application is literally the app that has all the code necessary to run the Erlang runtime system: file servers, code servers, and so on. She is the first application started. She's mandatory in the sense that the minimal system based on Erlang/OTP consists of Kernel and STDLIB. Kernel contains the following functional areas:

  • Start, stop, supervision, configuration, and distribution of applications
  • Code loading
  • Logging
  • Global name service
  • Supervision of Erlang/OTP
  • Communication with sockets
  • Operating system interface

🦋 LOGGER HANDLERS 🦋

Two standard logger handlers are defined in the Kernel application. These are described in the Kernel User's Guide, and in the logger_std_h(3erl) and logger_disk_log_h(3erl) manual pages. Obsessed. ✨

✨ OS SIGNAL EVENT HANDLER ✨

Asynchronous OS signals may be subscribed to via the Kernel application's event manager (see OTP Design Principles and gen_event(3erl)) registered as erl_signal_server. A default signal handler is installed which handles the following signals:

  • sigusr1 — The default handler will halt Erlang and produce a crashdump with slogan "Received SIGUSR1". This is equivalent to calling erlang:halt("Received SIGUSR1").
  • sigquit — The default handler will halt Erlang immediately. This is equivalent to calling erlang:halt().
  • sigterm — The default handler will terminate Erlang normally. This is equivalent to calling init:stop().

Events

Any event handler added to erl_signal_server must handle the following events.

  • sighup — Hangup detected on controlling terminal or death of controlling process
  • sigquit — Quit from keyboard
  • sigabrt — Abort signal from abort
  • sigalrm — Timer signal from alarm
  • sigterm — Termination signal
  • sigusr1 — User-defined signal 1
  • sigusr2 — User-defined signal 2
  • sigchld — Child process stopped or terminated
  • sigstop — Stop process
  • sigtstp — Stop typed at terminal

Setting OS signals are described in os:set_signal/2.

👑 CONFIGURATION 👑

The following configuration parameters are defined for the Kernel application. For more information about configuration parameters, see file app(5).

  • connect_all = true | false — If enabled (true), which also is the default, global(3erl) will actively connect to all nodes that becomes known to it. Note that you also want to enable prevent_overlapping_partitions in order for global to ensure that a fully connected network is maintained. prevent_overlapping_partitions will also prevent inconsistencies in global's name registration and locking.

The now deprecated command line argument -connect_all <boolean> has the same effect as the connect_all configuration parameter. If this configuration parameter is defined, it will override the command line argument.

  • distributed = [Distrib] — Specifies which applications that are distributed and on which nodes they are allowed to execute. In this parameter:
    • Distrib = {App,Nodes} | {App,Time,Nodes}
    • App = atom()
    • Time = integer()>0
    • Nodes = [node() | {node(),...,node()}]

The parameter is described in application:load/2.

  • dist_auto_connect = Value — Specifies when nodes are automatically connected. If this parameter is not specified, a node is always automatically connected, for example, when a message is to be sent to that node. Value is one of:
    • never — Connections are never automatically established, they must be explicitly connected. See net_kernel(3erl).
    • once — Connections are established automatically, but only once per node. If a node goes down, it must thereafter be explicitly connected. See net_kernel(3erl).
  • permissions = [Perm] — Specifies the default permission for applications when they are started. In this parameter:
    • Perm = {ApplName,Bool}
    • ApplName = atom()
    • Bool = boolean()

Permissions are described in application:permit/2.

  • logger = [Config] — Specifies the configuration for Logger, except the primary log level, which is specified with logger_level, and the compatibility with SASL Error Logging, which is specified with logger_sasl_compatible.

The logger parameter is described in section Logging in the Kernel User's Guide.

  • logger_level = Level — Specifies the primary log level for Logger. Log events with the same, or a more severe level, pass through the primary log level check. See section Logging in the Kernel User's Guide for more information about Logger and log levels.

Level = emergency | alert | critical | error | warning | notice | info | debug | all | none

To change the primary log level at runtime, use logger:set_primary_config(level, Level).

Defaults to notice.

  • logger_metadata = Metadata — Specifies primary metadata for log events.

Metadata = map()

Defaults to #{}.

  • logger_sasl_compatible = true | false — Specifies if Logger behaves backwards compatible with the SASL error logging functionality from releases prior to Erlang/OTP 21.0.

If this parameter is set to true, the default Logger handler does not log any progress-, crash-, or supervisor reports. If the SASL application is then started, it adds a Logger handler named sasl, which logs these events according to values of the SASL configuration parameter sasl_error_logger and sasl_errlog_type.

See section Deprecated Error Logger Event Handlers and Configuration in the sasl(7) manual page for information about the SASL configuration parameters.

See section SASL Error Logging in the SASL User's Guide, and section Backwards Compatibility with error_logger in the Kernel User's Guide for information about the SASL error logging functionality, and how Logger can be backwards compatible with this.

Defaults to false.

Note: If this parameter is set to true, sasl_errlog_type indicates that progress reports shall be logged, and the configured primary log level is notice or more severe, then SASL automatically sets the primary log level to info. That is, this setting can potentially overwrite the value of the Kernel configuration parameter logger_level. This is to allow progress reports, which have log level info, to be forwarded to the handlers.
  • global_groups = [GroupTuple] — Defines global groups, see global_group(3erl). In this parameter:
    • GroupTuple = {GroupName, [Node]} | {GroupName, PublishType, [Node]}
    • GroupName = atom()
    • PublishType = normal | hidden
    • Node = node()
  • inet_default_connect_options = [{Opt, Val}] — Specifies default options for connect sockets, see inet(3erl).
  • inet_default_listen_options = [{Opt, Val}] — Specifies default options for listen (and accept) sockets, see inet(3erl).
  • inet_dist_use_interface = ip_address() — If the host of an Erlang node has many network interfaces, this parameter specifies which one to listen on. For the type definition of ip_address(), see inet(3erl).
  • inet_dist_listen_min = First, inet_dist_listen_max = Last — Defines the First..Last port range for the listener socket of a distributed Erlang node.
  • inet_dist_listen_options = Opts — Defines a list of extra socket options to be used when opening the listening socket for a distributed Erlang node. See gen_tcp:listen/2.
  • inet_dist_connect_options = Opts — Defines a list of extra socket options to be used when connecting to other distributed Erlang nodes. See gen_tcp:connect/4.
  • inet_parse_error_log = silent — If set, no log events are issued when erroneous lines are found and skipped in the various Inet configuration files.
  • inetrc = Filename — The name (string) of an Inet user configuration file. For details, see section Inet Configuration in the ERTS User's Guide.
  • net_setuptime = SetupTime — SetupTime must be a positive integer or floating point number, and is interpreted as the maximum allowed time for each network operation during connection setup to another Erlang node. The maximum allowed value is 120. If higher values are specified, 120 is used. Default is 7 seconds if the variable is not specified, or if the value is incorrect (for example, not a number).

Notice that this value does not limit the total connection setup time, but rather each individual network operation during the connection setup and handshake.

  • net_ticker_spawn_options = Opts — Defines a list of extra spawn options for net ticker processes. There exist one such process for each connection to another node. A net ticker process is responsible for supervising the connection it is associated with. These processes also execute the distribution handshake protocol when setting up connections. When there is a large number of distribution connections, setting up garbage collection options can be helpful to reduce memory usage. Default is [link, {priority, max}], and these two options cannot be changed. The monitor and {monitor, MonitorOpts} options are not allowed and will be dropped if present. See the documentation of the erlang:spawn_opt/4 BIF for information about valid options. If the Opts list is not a proper list, or containing invalid options the setup of connections will fail.

Note that the behavior described above is only true if the distribution carrier protocol used is implemented as described in ERTS User's Guide ➜ How to implement an Alternative Carrier for the Erlang Distribution ➜ Distribution Module without further alterations. The implementer of the distribution carrier protocol used, may have chosen to ignore the net_ticker_spawn_options parameter or altered its behavior. Currently all distribution modules shipped with OTP do, however, behave as described above.

  • net_tickintensity = NetTickIntensity — okay so net tick intensity is literally how many ticks get sent during one net tick time period when there's no other data going over a connection to another node. It also decides how often we peek for data from that other node. The higher the intensity, the closer to your chosen net tick time period the node will detect an unresponsive node. Defaults to 4. And NetTickIntensity has to be an integer in the range 4..1000. Not an integer, or an integer less than 4? Then 4 gets silently used, no drama. Integer bigger than 1000? Then 1000 gets silently used. slay.
Note: all communicating nodes are expected to use the same net tick intensity and the same net tick time. matching fits, bestie 💖
Warning: do NOT set a too-high net tick intensity. You can genuinely overwhelm the node with work if it's set too high. 🫠
  • net_ticktime = NetTickTime — sets the net tick time in seconds. It's the approximate time a connected node may be unresponsive until it's considered down and thereby disconnected.

Net tick time together with net tick intensity determines an interval TickInterval = NetTickTime/NetTickIntensity. Once every TickInterval seconds, each connected node gets ticked if nothing has been sent to it during that last TickInterval seconds. And a tick? A tick is a small package sent on the connection. A connected node is considered to be down if no ticks or payload packages have been received during the last NetTickIntensity number of TickInterval seconds intervals. This ensures that nodes that aren't responding — for reasons such as hardware errors — are considered to be down.

As the availability is only checked every TickInterval seconds, the actual time T a node has been unresponsive when detected may vary between MinT and MaxT, where:

MinT = NetTickTime - NetTickTime / NetTickIntensity
MaxT = NetTickTime + NetTickTime / NetTickIntensity

NetTickTime defaults to 60 seconds and NetTickIntensity defaults to 4. Thus, 45 < T < 75 seconds. the math is mathing 💅

Note: notice that all communicating nodes are to have the same NetTickTime and NetTickIntensity values specified, as it determines both the frequency of outgoing ticks and the expected frequency of incoming ticks.

NetTickTime needs to be a multiple of NetTickIntensity. If the configured values aren't, NetTickTime will internally be rounded up to the nearest millisecond. net_kernel:get_net_ticktime() will, however, report net tick time truncated to the nearest second.

Normally, a terminating node is detected immediately by the transport protocol (like TCP/IP).

  • prevent_overlapping_partitions = true | false — if enabled (true), global will actively prevent overlapping partitions from forming when connections are lost between nodes. This fix is enabled by default. If you're about to disable this fix, make sure to read the global(3erl) documentation about this fix for more important information about this. fr fr, go read it first 🙈
  • shutdown_timeout = integer() | infinity — specifies the time application_controller waits for an application to terminate during node shutdown. If the timer expires, application_controller brutally kills application_master of the hanging application. If this parameter is undefined, it defaults to infinity.
  • sync_nodes_mandatory = [NodeName] — specifies which other nodes must be alive for this node to start properly. If some node in the list doesn't start within the specified time, this node doesn't start either. If this parameter is undefined, it defaults to [].
  • sync_nodes_optional = [NodeName] — specifies which other nodes can be alive for this node to start properly. If some node in this list doesn't start within the specified time, this node starts anyway. If this parameter is undefined, it defaults to the empty list.
  • sync_nodes_timeout = integer() | infinity — specifies the time (in milliseconds) that this node waits for the mandatory and optional nodes to start. If this parameter is undefined, no node synchronization is performed. This option ensures that global is synchronized.
  • start_distribution = true | false — starts all distribution services, such as rpc, global, and net_kernel if the parameter is true. This parameter is to be set to false for systems who want to disable all distribution functionality.

Defaults to true.

  • start_dist_ac = true | false — starts the dist_ac server if the parameter is true. This parameter is to be set to true for systems using distributed applications.

Defaults to false. If this parameter is undefined, the server is started if parameter distributed is set.

  • start_boot_server = true | false — starts the boot_server if the parameter is true (see erl_boot_server(3erl)). This parameter is to be set to true in an embedded system using this service.

Defaults to false.

  • boot_server_slaves = [SlaveIP] — if configuration parameter start_boot_server is true, this parameter can be used to initialize boot_server with a list of slave IP addresses:
SlaveIP = string() | atom |
{integer(),integer(),integer(),integer()},

where 0 <= integer() <=255.

Examples of SlaveIP in atom, string, and tuple form:

'150.236.16.70', "150,236,16,70", {150,236,16,70}.

Defaults to [].

  • start_disk_log = true | false — starts the disk_log_server if the parameter is true (see disk_log(3erl)). This parameter is to be set to true in an embedded system using this service.

Defaults to false.

  • start_pg = true | false — starts the default pg scope server (see pg(3erl)) if the parameter is true. This parameter is to be set to true in an embedded system that uses this service.

Defaults to false.

  • start_timer = true | false — starts the timer_server if the parameter is true (see timer(3erl)). This parameter is to be set to true in an embedded system using this service.

Defaults to false.

  • shell_history = enabled | disabled | module() — specifies whether shell history should be logged to disk between usages of erl (enabled), not logged at all (disabled), or a user-specified module will be used to log shell history. This module should export load() -> [string()] returning a list of strings to load in the shell when it starts, and add(iodata()) -> ok called every time a new line is entered in the shell. By default logging is disabled.
  • shell_history_drop = [string()] — specific log lines that should not be persisted. For example ["q().", "init:stop()."] will allow you to ignore commands that shut the node down. Defaults to [].
  • shell_history_file_bytes = integer() — how many bytes the shell should remember. By default, the value is set to 512kb, and the minimal value is 50kb.
  • shell_history_path = string() — specifies where the shell history files will be stored. Defaults to the user's cache directory as returned by filename:basedir(user_cache, "erlang-history").
  • shutdown_func = {Mod, Func} — where:
    • Mod = atom()
    • Func = atom()

Sets a function that application_controller calls when it starts to terminate. The function is called as Mod:Func(Reason), where Reason is the terminate reason for application_controller, and it must return as soon as possible for application_controller to terminate properly. no dawdling in there bestie 🏃‍♀️

  • source_search_rules = [DirRule] | [SuffixRule] — where:
    • DirRule = {ObjDirSuffix,SrcDirSuffix}
    • SuffixRule = {ObjSuffix,SrcSuffix,[DirRule]}
    • ObjDirSuffix = string()
    • SrcDirSuffix = string()
    • ObjSuffix = string()
    • SrcSuffix = string()

Specifies a list of rules for use by filelib:find_file/2 and filelib:find_source/2. If this is set to some other value than the empty list, it replaces the default rules. Rules can be simple pairs of directory suffixes, such as {"ebin", "src"}, which are used by filelib:find_file/2, or triples specifying separate directory suffix rules depending on file name extensions, for example [{".beam", ".erl", [{"ebin", "src"}]}, which are used by filelib:find_source/2. Both kinds of rules can be mixed in the list.

The interpretation of ObjDirSuffix and SrcDirSuffix is as follows: if the end of the directory name where an object is located matches ObjDirSuffix, then the name created by replacing ObjDirSuffix with SrcDirSuffix is expanded by calling filelib:wildcard/1, and the first regular file found among the matches is the source file.

💀 DEPRECATED CONFIGURATION PARAMETERS 💀

In Erlang/OTP 21.0, a new API for logging was added. The old error_logger event manager, and event handlers running on this manager, still work, but they are no longer used by default. it's giving legacy, and not in a cute way 😭

The following application configuration parameters can still be set, but they are only used if the corresponding configuration parameters for Logger are not set.

  • error_logger — replaced by setting the type, and possibly file and modes parameters of the default logger_std_h handler. Example:
erl -kernel logger '[{handler,default,logger_std_h,#{config=>#{file=>"/tmp/erlang.log"}}}]'
  • error_logger_format_depth — replaced by setting the depth parameter of the default handler's formatter. Example:
erl -kernel logger '[{handler,default,logger_std_h,#{formatter=>{logger_formatter,#{legacy_header=>true,template=>[{logger_formatter,header},"\n",msg,"\n"],depth=>10}}}]'

See Backwards compatibility with error_logger for more information.

✨ SEE ALSO ✨

app(5), application(3erl), code(3erl), disk_log(3erl), erl_boot_server(3erl), erl_ddll(3erl), file(3erl), global(3erl), global_group(3erl), heart(3erl), inet(3erl), logger(3erl), net_kernel(3erl), os(3erl), pg(3erl), rpc(3erl), seq_trace(3erl), user(3erl), timer(3erl)

and that's the whole page, bestie 🫶 this one came from Ericsson AB, version kernel 8.5.4.2, and the date stamp lives down there in the footer of your local copy — go check it, i'll wait 💅✨🎀

📦 where this page comes from