nixpkgs/nixos/modules/services/search/elasticsearch.nix
talyz 3a1e1f0624
nixos/elasticsearch: Wait for elasticsearch to start up properly
Other services that depend on elasticsearch should be started after
it, but since the versions we're packaging have to run as
"Type=simple", they're started as soon as the elasticsearch binary has
been executed, likely winning the race against it.

This makes sure elasticsearch is up and running, responding to a
simple query, before dependents are started.
2021-09-13 13:57:01 +02:00

225 lines
6.5 KiB
Nix

{ config, lib, pkgs, ... }:
with lib;
let
cfg = config.services.elasticsearch;
esConfig = ''
network.host: ${cfg.listenAddress}
cluster.name: ${cfg.cluster_name}
${lib.optionalString cfg.single_node ''
discovery.type: single-node
gateway.auto_import_dangling_indices: true
''}
http.port: ${toString cfg.port}
transport.port: ${toString cfg.tcp_port}
${cfg.extraConf}
'';
configDir = cfg.dataDir + "/config";
elasticsearchYml = pkgs.writeTextFile {
name = "elasticsearch.yml";
text = esConfig;
};
loggingConfigFilename = "log4j2.properties";
loggingConfigFile = pkgs.writeTextFile {
name = loggingConfigFilename;
text = cfg.logging;
};
esPlugins = pkgs.buildEnv {
name = "elasticsearch-plugins";
paths = cfg.plugins;
postBuild = "${pkgs.coreutils}/bin/mkdir -p $out/plugins";
};
in
{
###### interface
options.services.elasticsearch = {
enable = mkOption {
description = "Whether to enable elasticsearch.";
default = false;
type = types.bool;
};
package = mkOption {
description = "Elasticsearch package to use.";
default = pkgs.elasticsearch;
defaultText = "pkgs.elasticsearch";
type = types.package;
};
listenAddress = mkOption {
description = "Elasticsearch listen address.";
default = "127.0.0.1";
type = types.str;
};
port = mkOption {
description = "Elasticsearch port to listen for HTTP traffic.";
default = 9200;
type = types.int;
};
tcp_port = mkOption {
description = "Elasticsearch port for the node to node communication.";
default = 9300;
type = types.int;
};
cluster_name = mkOption {
description = "Elasticsearch name that identifies your cluster for auto-discovery.";
default = "elasticsearch";
type = types.str;
};
single_node = mkOption {
description = "Start a single-node cluster";
default = true;
type = types.bool;
};
extraConf = mkOption {
description = "Extra configuration for elasticsearch.";
default = "";
type = types.str;
example = ''
node.name: "elasticsearch"
node.master: true
node.data: false
'';
};
logging = mkOption {
description = "Elasticsearch logging configuration.";
default = ''
logger.action.name = org.elasticsearch.action
logger.action.level = info
appender.console.type = Console
appender.console.name = console
appender.console.layout.type = PatternLayout
appender.console.layout.pattern = [%d{ISO8601}][%-5p][%-25c{1.}] %marker%m%n
rootLogger.level = info
rootLogger.appenderRef.console.ref = console
'';
type = types.str;
};
dataDir = mkOption {
type = types.path;
default = "/var/lib/elasticsearch";
description = ''
Data directory for elasticsearch.
'';
};
extraCmdLineOptions = mkOption {
description = "Extra command line options for the elasticsearch launcher.";
default = [ ];
type = types.listOf types.str;
};
extraJavaOptions = mkOption {
description = "Extra command line options for Java.";
default = [ ];
type = types.listOf types.str;
example = [ "-Djava.net.preferIPv4Stack=true" ];
};
plugins = mkOption {
description = "Extra elasticsearch plugins";
default = [ ];
type = types.listOf types.package;
example = lib.literalExample "[ pkgs.elasticsearchPlugins.discovery-ec2 ]";
};
};
###### implementation
config = mkIf cfg.enable {
systemd.services.elasticsearch = {
description = "Elasticsearch Daemon";
wantedBy = [ "multi-user.target" ];
after = [ "network.target" ];
path = [ pkgs.inetutils ];
environment = {
ES_HOME = cfg.dataDir;
ES_JAVA_OPTS = toString cfg.extraJavaOptions;
ES_PATH_CONF = configDir;
};
serviceConfig = {
ExecStart = "${cfg.package}/bin/elasticsearch ${toString cfg.extraCmdLineOptions}";
User = "elasticsearch";
PermissionsStartOnly = true;
LimitNOFILE = "1024000";
};
preStart = ''
${optionalString (!config.boot.isContainer) ''
# Only set vm.max_map_count if lower than ES required minimum
# This avoids conflict if configured via boot.kernel.sysctl
if [ `${pkgs.procps}/bin/sysctl -n vm.max_map_count` -lt 262144 ]; then
${pkgs.procps}/bin/sysctl -w vm.max_map_count=262144
fi
''}
mkdir -m 0700 -p ${cfg.dataDir}
# Install plugins
ln -sfT ${esPlugins}/plugins ${cfg.dataDir}/plugins
ln -sfT ${cfg.package}/lib ${cfg.dataDir}/lib
ln -sfT ${cfg.package}/modules ${cfg.dataDir}/modules
# elasticsearch needs to create the elasticsearch.keystore in the config directory
# so this directory needs to be writable.
mkdir -m 0700 -p ${configDir}
# Note that we copy config files from the nix store instead of symbolically linking them
# because otherwise X-Pack Security will raise the following exception:
# java.security.AccessControlException:
# access denied ("java.io.FilePermission" "/var/lib/elasticsearch/config/elasticsearch.yml" "read")
cp ${elasticsearchYml} ${configDir}/elasticsearch.yml
# Make sure the logging configuration for old elasticsearch versions is removed:
rm -f "${configDir}/logging.yml"
cp ${loggingConfigFile} ${configDir}/${loggingConfigFilename}
mkdir -p ${configDir}/scripts
cp ${cfg.package}/config/jvm.options ${configDir}/jvm.options
# redirect jvm logs to the data directory
mkdir -m 0700 -p ${cfg.dataDir}/logs
${pkgs.sd}/bin/sd 'logs/gc.log' '${cfg.dataDir}/logs/gc.log' ${configDir}/jvm.options \
if [ "$(id -u)" = 0 ]; then chown -R elasticsearch:elasticsearch ${cfg.dataDir}; fi
'';
postStart = ''
# Make sure elasticsearch is up and running before dependents
# are started
while ! ${pkgs.curl}/bin/curl -sS -f http://localhost:${toString cfg.port} 2>/dev/null; do
sleep 1
done
'';
};
environment.systemPackages = [ cfg.package ];
users = {
groups.elasticsearch.gid = config.ids.gids.elasticsearch;
users.elasticsearch = {
uid = config.ids.uids.elasticsearch;
description = "Elasticsearch daemon user";
home = cfg.dataDir;
group = "elasticsearch";
};
};
};
}