1 class Node < ArvadosModel
4 include CommonApiTemplate
6 before_validation :ensure_ping_secret
7 after_update :dnsmasq_update
11 @@confdir = if Rails.configuration.respond_to? :dnsmasq_conf_dir
12 Rails.configuration.dnsmasq_conf_dir
13 elsif File.exists? '/etc/dnsmasq.d/.'
18 @@domain = Rails.configuration.compute_node_domain rescue `hostname --domain`.strip
19 @@nameservers = Rails.configuration.compute_node_nameservers
21 api_accessible :user, :extend => :common do |t|
29 api_accessible :superuser, :extend => :user do |t|
32 t.add lambda { |x| @@nameservers }, :as => :nameservers
46 if Time.now - self.created_at > 5.minutes
51 elsif Time.now - self.last_ping_at > 1.hours
59 raise "must have :ip and :ping_secret" unless o[:ip] and o[:ping_secret]
61 if o[:ping_secret] != self.info[:ping_secret]
62 logger.info "Ping: secret mismatch: received \"#{o[:ping_secret]}\" != \"#{self.info[:ping_secret]}\""
65 self.last_ping_at = Time.now
67 @bypass_arvados_authorization = true
70 if self.ip_address.nil?
71 logger.info "#{self.uuid} ip_address= #{o[:ip]}"
72 self.ip_address = o[:ip]
73 self.first_ping_at = Time.now
76 # Record instance ID if not already known
77 if o[:ec2_instance_id]
78 if !self.info[:ec2_instance_id]
79 self.info[:ec2_instance_id] = o[:ec2_instance_id]
80 `ec2-create-tags #{o[:ec2_instance_id]} --tag 'Name=#{self.uuid}'`
81 elsif self.info[:ec2_instance_id] != o[:ec2_instance_id]
82 logger.debug "Multiple nodes have credentials for #{self.uuid}"
83 raise "#{self.uuid} is already running at #{self.info[:ec2_instance_id]} so rejecting ping from #{o[:ec2_instance_id]}"
88 if self.slot_number.nil?
91 self.slot_number = try_slot
95 rescue ActiveRecord::RecordNotUnique
98 raise "No available node slots" if try_slot == MAX_SLOTS
100 self.hostname = self.class.hostname_for_slot(self.slot_number)
101 if info[:ec2_instance_id]
102 `ec2-create-tags #{self.info[:ec2_instance_id]} --tag 'hostname=#{self.hostname}'`
109 def start!(ping_url_method)
110 ensure_permission_to_update
111 ping_url = ping_url_method.call({ uuid: self.uuid, ping_secret: self.info[:ping_secret] })
112 ec2_args = ["--user-data '#{ping_url}'",
114 Rails.configuration.compute_node_ec2run_args,
115 Rails.configuration.compute_node_ami
117 ec2run_cmd = ["ec2-run-instances",
118 "--client-token", self.uuid,
119 ec2_args].flatten.join(' ')
120 ec2spot_cmd = ["ec2-request-spot-instances",
121 "-p #{Rails.configuration.compute_node_spot_bid} --type one-time",
122 ec2_args].flatten.join(' ')
123 self.info[:ec2_run_command] = ec2run_cmd
124 self.info[:ec2_spot_command] = ec2spot_cmd
125 self.info[:ec2_start_command] = ec2spot_cmd
126 logger.info "#{self.uuid} ec2_start_command= #{ec2spot_cmd.inspect}"
127 result = `#{ec2spot_cmd} 2>&1`
128 self.info[:ec2_start_result] = result
129 logger.info "#{self.uuid} ec2_start_result= #{result.inspect}"
130 result.match(/INSTANCE\s*(i-[0-9a-f]+)/) do |m|
132 self.info[:ec2_instance_id] = instance_id
133 `ec2-create-tags #{instance_id} --tag 'Name=#{self.uuid}'`
135 result.match(/SPOTINSTANCEREQUEST\s*(sir-[0-9a-f]+)/) do |m|
137 self.info[:ec2_sir_id] = sir_id
138 `ec2-create-tags #{sir_id} --tag 'Name=#{self.uuid}'`
145 def ensure_ping_secret
146 self.info[:ping_secret] ||= rand(2**256).to_s(36)
150 if self.hostname_changed? or self.ip_address_changed?
151 if self.hostname and self.ip_address
152 self.class.dnsmasq_update(self.hostname, self.ip_address)
157 def self.dnsmasq_update(hostname, ip_address)
158 return unless @@confdir
159 ptr_domain = ip_address.
160 split('.').reverse.join('.').concat('.in-addr.arpa')
161 hostfile = File.join @@confdir, hostname
162 File.open hostfile, 'w' do |f|
163 f.puts "address=/#{hostname}/#{ip_address}"
164 f.puts "address=/#{hostname}.#{@@domain}/#{ip_address}" if @@domain
165 f.puts "ptr-record=#{ptr_domain},#{hostname}"
167 File.open(File.join(@@confdir, 'restart.txt'), 'w') do |f|
168 # this should trigger a dnsmasq restart
172 def self.hostname_for_slot(slot_number)
173 "compute#{slot_number}"
176 # At startup, make sure all DNS entries exist. Otherwise, slurmctld
177 # will refuse to start.
179 !File.exists? (File.join(@@confdir, hostname_for_slot(MAX_SLOTS-1)))
180 (0..MAX_SLOTS-1).each do |slot_number|
181 hostname = hostname_for_slot(slot_number)
182 hostfile = File.join @@confdir, hostname
183 if !File.exists? hostfile
184 dnsmasq_update(hostname, '127.40.4.0')
189 def permission_to_update
190 @bypass_arvados_authorization or super
193 def permission_to_create
194 current_user and current_user.is_admin