X-Git-Url: https://git.openstreetmap.org./chef.git/blobdiff_plain/df246797e4b8705a0912158c7aa46e4da3138e01..794cc957324de66edab45373053be3e601090f1f:/cookbooks/nominatim/recipes/default.rb?ds=inline diff --git a/cookbooks/nominatim/recipes/default.rb b/cookbooks/nominatim/recipes/default.rb index 75a893841..7eb6a1e74 100644 --- a/cookbooks/nominatim/recipes/default.rb +++ b/cookbooks/nominatim/recipes/default.rb @@ -18,11 +18,19 @@ # include_recipe "accounts" -include_recipe "munin" -include_recipe "php::fpm" +include_recipe "prometheus" +include_recipe "postgresql" +include_recipe "python" +include_recipe "nginx" +include_recipe "git" +include_recipe "fail2ban" basedir = data_bag_item("accounts", "nominatim")["home"] -email_errors = data_bag_item("accounts", "lonvia")["email"] +project_directory = "#{basedir}/planet-project" +bin_directory = "#{basedir}/bin" +cfg_directory = "#{basedir}/etc" +ui_directory = "#{basedir}/ui" +qa_data_directory = "#{basedir}/qa-data" directory basedir do owner "nominatim" @@ -31,6 +39,28 @@ directory basedir do recursive true end +[basedir, bin_directory, cfg_directory, project_directory, ui_directory].each do |path| + directory path do + owner "nominatim" + group "nominatim" + mode "755" + end +end + +if node[:nominatim][:flatnode_file] + directory File.dirname(node[:nominatim][:flatnode_file]) do + recursive true + end +end + +directory "#{bin_directory}/maintenance" do + owner "nominatim" + group "nominatim" + mode "775" +end + +## Log directory setup + directory node[:nominatim][:logdir] do owner "nominatim" group "nominatim" @@ -45,27 +75,7 @@ file "#{node[:nominatim][:logdir]}/query.log" do mode "664" end -file "#{node[:nominatim][:logdir]}/update.log" do - action :create_if_missing - owner "nominatim" - group "adm" - mode "664" -end - -# exception granted for a limited time so that they can set up their own server -firewall_rule "increase-limits-gnome-proxy" do - action :accept - family "inet" - source "net:8.43.85.23" - dest "fw" - proto "tcp:syn" - dest_ports "https" - rate_limit "s:10/sec:30" -end - -## Postgresql - -include_recipe "postgresql" +### Postgresql postgresql_version = node[:nominatim][:dbcluster].split("/").first postgis_version = node[:nominatim][:postgis] @@ -76,302 +86,104 @@ node[:nominatim][:dbadmins].each do |user| postgresql_user user do cluster node[:nominatim][:dbcluster] superuser true - only_if { node[:nominatim][:state] != "slave" } end end postgresql_user "nominatim" do cluster node[:nominatim][:dbcluster] superuser true - only_if { node[:nominatim][:state] != "slave" } end postgresql_user "www-data" do cluster node[:nominatim][:dbcluster] - only_if { node[:nominatim][:state] != "slave" } -end - -postgresql_munin "nominatim" do - cluster node[:nominatim][:dbcluster] - database node[:nominatim][:dbname] -end - -directory "#{basedir}/tablespaces" do - owner "postgres" - group "postgres" - mode "700" -end - -# Note: tablespaces must be exactly in the same location on each -# Nominatim instance when replication is in use. Therefore -# use symlinks to canonical directory locations. -node[:nominatim][:tablespaces].each do |name, location| - directory location do - owner "postgres" - group "postgres" - mode "700" - recursive true - end - - link "#{basedir}/tablespaces/#{name}" do - to location - end - - postgresql_tablespace name do - cluster node[:nominatim][:dbcluster] - location "#{basedir}/tablespaces/#{name}" - end -end - -if node[:nominatim][:state] == "master" - postgresql_user "replication" do - cluster node[:nominatim][:dbcluster] - password data_bag_item("nominatim", "passwords")["replication"] - replication true - end - - directory node[:rsyncd][:modules][:archive][:path] do - owner "postgres" - group "postgres" - mode "700" - end - - template "/usr/local/bin/clean-db-nominatim" do - source "clean-db-nominatim.erb" - owner "root" - group "root" - mode "755" - variables :archive_dir => node[:rsyncd][:modules][:archive][:path], - :update_stop_file => "#{basedir}/status/updates_disabled", - :streaming_clients => search(:node, "nominatim_state:slave").map { |slave| slave[:fqdn] }.join(" ") - end end -## Nominatim backend +### Nominatim -include_recipe "git" +python_directory = "#{basedir}/venv" package %w[ build-essential - cmake - g++ - libboost-dev - libboost-system-dev - libboost-filesystem-dev - libexpat1-dev - zlib1g-dev - libxml2-dev - libbz2-dev - libpq-dev - libgeos++-dev - libproj-dev - python3-pyosmium - pyosmium - python3-psycopg2 - php-pgsql - php-intl + libicu-dev + python3-dev + pkg-config + osm2pgsql + ruby + ruby-file-tail + ruby-pg + ruby-webrick ] -source_directory = "#{basedir}/nominatim" -build_directory = "#{basedir}/bin" -ui_directory = "#{basedir}/ui" - -directory build_directory do - owner "nominatim" - group "nominatim" - mode "755" - recursive true +python_virtualenv python_directory do + interpreter "/usr/bin/python3" end -# Normally syncing via chef is a bad idea because syncing might involve -# an update of database functions which should not be done while an update -# is ongoing. Therefore we sync in between update cycles. There is an -# exception for slaves: they get DB function updates from the master, so -# only the source code needs to be updated, which chef may do. -git source_directory do - action node[:nominatim][:state] == "slave" ? :sync : :checkout - repository node[:nominatim][:repository] - revision node[:nominatim][:revision] - enable_submodules true - user "nominatim" - group "nominatim" - not_if { node[:nominatim][:state] != "slave" && File.exist?("#{source_directory}/README.md") } - notifies :run, "execute[compile_nominatim]", :immediately +# These are updated during the database update. +python_package "nominatim-db" do + python_virtualenv python_directory + extra_index_url node[:nominatim][:pip_index] end -execute "compile_nominatim" do - action :nothing - user "nominatim" - cwd build_directory - command "cmake #{source_directory} && make" +python_package "nominatim-api" do + python_virtualenv python_directory + extra_index_url node[:nominatim][:pip_index] end -template "#{source_directory}/.git/hooks/post-merge" do - source "git-post-merge-hook.erb" +remote_directory "#{project_directory}/static-website" do + source "website" owner "nominatim" group "nominatim" mode "755" - variables :srcdir => source_directory, - :builddir => build_directory, - :dbname => node[:nominatim][:dbname] + files_owner "nominatim" + files_group "nominatim" + files_mode "644" + purge false end -template "#{build_directory}/settings/local.php" do - source "settings.erb" +template "#{project_directory}/.env" do + source "nominatim.env.erb" owner "nominatim" group "nominatim" mode "664" - variables :base_url => node[:nominatim][:state] == "off" ? node[:fqdn] : "nominatim.openstreetmap.org", + variables :base_url => "nominatim.openstreetmap.org", :dbname => node[:nominatim][:dbname], :flatnode_file => node[:nominatim][:flatnode_file], - :log_file => "#{node[:nominatim][:logdir]}/query.log" -end - -git ui_directory do - action :sync - repository node[:nominatim][:ui_repository] - revision node[:nominatim][:ui_revision] - user "nominatim" - group "nominatim" + :log_file => "#{node[:nominatim][:logdir]}/query.log", + :pool_size => node[:nominatim][:api_pool_size], + :query_timeout => node[:nominatim][:api_query_timeout], + :request_timeout => node[:nominatim][:api_request_timeout] end -template "#{ui_directory}/dist/config.js" do - source "ui-config.js.erb" +remote_file "#{project_directory}/secondary_importance.sql.gz" do + action :create_if_missing + source "https://nominatim.org/data/wikimedia-secondary-importance.sql.gz" owner "nominatim" group "nominatim" - mode 0o664 -end - -if node[:nominatim][:flatnode_file] - directory File.dirname(node[:nominatim][:flatnode_file]) do - recursive true - end -end - -template "/etc/logrotate.d/nominatim" do - source "logrotate.nominatim.erb" - owner "root" - group "root" mode "644" end -external_data = [ - "wikimedia-importance.sql.gz", - "gb_postcode_data.sql.gz" -] - -external_data.each do |fname| - remote_file "#{source_directory}/data/#{fname}" do - action :create_if_missing - source "https://www.nominatim.org/data/#{fname}" - owner "nominatim" - group "nominatim" - mode "644" - end -end - -remote_file "#{source_directory}/data/country_osm_grid.sql.gz" do +remote_file "#{project_directory}/wikimedia-importance.csv.gz" do action :create_if_missing - source "https://www.nominatim.org/data/country_grid.sql.gz" + source "https://nominatim.org/data/wikimedia-importance.csv.gz" owner "nominatim" group "nominatim" mode "644" end -if node[:nominatim][:state] == "off" - cron_d "nominatim-backup" do - action :delete - end - - cron_d "nominatim-vacuum-db" do - action :delete - end - - cron_d "nominatim-clean-db" do - action :delete - end - - cron_d "nominatim-update-maintenance-trigger" do - action :delete - end -else - cron_d "nominatim-backup" do - action node[:nominatim][:enable_backup] ? :create : :delete - minute "0" - hour "3" - day "1" - user "nominatim" - command "/usr/local/bin/backup-nominatim" - mailto email_errors - end - - cron_d "nominatim-vacuum-db" do - minute "20" - hour "0" - user "postgres" - command "/usr/local/bin/vacuum-db-nominatim" - mailto email_errors - end - - cron_d "nominatim-clean-db" do - action node[:nominatim][:state] == "master" ? :create : :delete - minute "5" - hour "*/4" - user "postgres" - command "/usr/local/bin/clean-db-nominatim" - mailto email_errors - end - - cron_d "nominatim-update-maintenance-trigger" do - minute "18" - hour "1" - user "nominatim" - command "touch #{basedir}/status/update_maintenance" - mailto email_errors - end -end - -template "#{source_directory}/utils/nominatim-update" do - source "updater.erb" - user "nominatim" - group "nominatim" - mode "755" - variables :bindir => build_directory, - :srcdir => source_directory, - :logfile => "#{node[:nominatim][:logdir]}/update.log", - :branch => node[:nominatim][:revision], - :update_stop_file => "#{basedir}/status/updates_disabled", - :update_maintenance_trigger => "#{basedir}/status/update_maintenance" -end - -template "/etc/init.d/nominatim-update" do - source "updater.init.erb" - user "nominatim" - group "nominatim" - mode "755" - variables :source_directory => source_directory -end - -%w[backup-nominatim vacuum-db-nominatim].each do |fname| - template "/usr/local/bin/#{fname}" do - source "#{fname}.erb" - owner "root" - group "root" - mode "755" - variables :db => node[:nominatim][:dbname] +%w[gb_postcodes.csv.gz us_postcodes.csv.gz].each do |fname| + remote_file "#{project_directory}/#{fname}" do + action :create + source "https://nominatim.org/data/#{fname}" + owner "nominatim" + group "nominatim" + mode "644" end end -## webserver frontend - -directory "#{basedir}/etc" do - owner "nominatim" - group "adm" - mode "775" -end +# Webserver + frontend -%w[user_agent referrer email].each do |name| - file "#{basedir}/etc/nginx_blocked_#{name}.conf" do +%w[user_agent referrer email generic].each do |name| + file "#{cfg_directory}/nginx_blocked_#{name}.conf" do action :create_if_missing owner "nominatim" group "adm" @@ -379,16 +191,26 @@ end end end -node[:nominatim][:fpm_pools].each do |name, data| - php_fpm name.to_s do - port data[:port] - pm data[:pm] - pm_max_children data[:max_children] - pm_start_servers 20 - pm_min_spare_servers 10 - pm_max_spare_servers 20 - pm_max_requests 10000 - end +systemd_service "nominatim" do + description "Nominatim running as a gunicorn application" + user "www-data" + group "www-data" + working_directory project_directory + standard_output "append:#{node[:nominatim][:logdir]}/gunicorn.log" + standard_error "inherit" + exec_start "#{python_directory}/bin/gunicorn --max-requests 200000 -b unix:/run/gunicorn-nominatim.openstreetmap.org.sock -w #{node[:nominatim][:api_workers]} -k uvicorn.workers.UvicornWorker 'nominatim_api.server.falcon.server:run_wsgi()'" + exec_reload "/bin/kill -s HUP $MAINPID" + kill_mode "mixed" + timeout_stop_sec 5 + private_tmp true + requires "nominatim.socket" + after "network.target" +end + +systemd_socket "nominatim" do + description "Gunicorn socket for Nominatim" + listen_stream "/run/gunicorn-nominatim.openstreetmap.org.sock" + socket_user "www-data" end ssl_certificate node[:fqdn] do @@ -398,25 +220,22 @@ ssl_certificate node[:fqdn] do "nominatim.openstreetmap.com", "nominatim.openstreetmap.net", "nominatim.openstreetmaps.org", - "nominatim.openmaps.org"] + "nominatim.openmaps.org", + "nominatim.qgis.org"] notifies :reload, "service[nginx]" end -package "apache2" do - action :remove -end - -include_recipe "nginx" - nginx_site "default" do action [:delete] end +frontends = search(:node, "recipes:web\\:\\:frontend").sort_by(&:name) + nginx_site "nominatim" do template "nginx.erb" - directory build_directory + directory project_directory variables :pools => node[:nominatim][:fpm_pools], - :frontends => search(:node, "recipes:web\\:\\:frontend"), + :frontends => frontends, :confdir => "#{basedir}/etc", :ui_directory => ui_directory end @@ -428,35 +247,126 @@ template "/etc/logrotate.d/nginx" do mode "644" end -munin_plugin_conf "nominatim" do - template "munin.erb" - variables :db => node[:nominatim][:dbname], - :querylog => "#{node[:nominatim][:logdir]}/query.log" +### Import, update and maintenance scripts + +%w[nominatim-update + nominatim-update-data + nominatim-update-refresh-db + nominatim-daily-maintenance].each do |fname| + template "#{bin_directory}/#{fname}" do + source "#{fname}.erb" + owner "nominatim" + group "nominatim" + mode "554" + variables :bindir => bin_directory, + :projectdir => project_directory, + :venvprefix => "#{python_directory}/bin/", + :qadatadir => qa_data_directory + end +end + +systemd_service "nominatim-update" do + description "Update the Nominatim database" + exec_start "#{bin_directory}/nominatim-update" + restart "on-success" + standard_output "journal" + standard_error "inherit" + working_directory project_directory +end + +systemd_service "nominatim-update-maintenance-trigger" do + description "Trigger daily maintenance tasks for Nominatim DB" + exec_start "ln -sf #{bin_directory}/nominatim-daily-maintenance #{bin_directory}/maintenance/" + user "nominatim" end -munin_plugin "nominatim_importlag" do - target "#{source_directory}/munin/nominatim_importlag" +systemd_timer "nominatim-update-maintenance-trigger" do + action :create + description "Schedule daily maintenance tasks for Nominatim DB" + on_calendar "*-*-* 02:03:00 UTC" end -munin_plugin "nominatim_query_speed" do - target "#{source_directory}/munin/nominatim_query_speed_querylog" +service "nominatim-update-maintenance-trigger" do + action :enable end -munin_plugin "nominatim_requests" do - target "#{source_directory}/munin/nominatim_requests_querylog" +## Nominatim UI + +git ui_directory do + action :sync + repository node[:nominatim][:ui_repository] + revision node[:nominatim][:ui_revision] + user "nominatim" + group "nominatim" end -directory "#{basedir}/status" do +template "#{ui_directory}/dist/theme/config.theme.js" do + source "ui-config.js.erb" owner "nominatim" - group "postgres" - mode "775" + group "nominatim" + mode "664" end -include_recipe "fail2ban" +## Nominatim QA + +if node[:nominatim][:enable_qa_tiles] + python_package "nominatim-data-analyser" do + python_virtualenv python_directory + extra_index_url node[:nominatim][:pip_index] + end + + directory qa_data_directory do + owner "nominatim" + group "nominatim" + mode "755" + recursive true + end + + template "#{project_directory}/qa-config.yaml" do + source "qa_config.erb" + owner "nominatim" + group "nominatim" + mode "755" + variables :outputdir => "#{qa_data_directory}/new" + end + + ssl_certificate "qa-tile.nominatim.openstreetmap.org" do + domains ["qa-tile.nominatim.openstreetmap.org"] + notifies :reload, "service[nginx]" + end + + nginx_site "qa-tiles.nominatim" do + template "nginx-qa-tiles.erb" + directory qa_data_directory + variables :qa_data_directory => qa_data_directory + end +end + +## Logging and monitoring + +template "/etc/logrotate.d/nominatim" do + source "logrotate.nominatim.erb" + owner "root" + group "root" + mode "644" +end + +prometheus_exporter "nominatim" do + port 8082 + user "www-data" + restrict_address_families "AF_UNIX" + options [ + "--nominatim.query-log=#{node[:nominatim][:logdir]}/query.log", + "--nominatim.database-name=#{node[:nominatim][:dbname]}" + ] +end + +frontend_addresses = frontends.collect { |f| f.ipaddresses(:role => :external) } fail2ban_jail "nominatim_limit_req" do filter "nginx-limit-req" logpath "#{node[:nominatim][:logdir]}/nominatim.openstreetmap.org-error.log" ports [80, 443] - maxretry 5 + maxretry 20 + ignoreips frontend_addresses.flatten.sort end