1 class Node < ArvadosModel
4 include CommonApiTemplate
6 serialize :properties, Hash
7 before_validation :ensure_ping_secret
8 after_update :dns_server_update
10 # Only a controller can figure out whether or not the current API tokens
11 # have access to the associated Job. They're expected to set
12 # job_readable=true if the Job UUID can be included in the API response.
13 belongs_to(:job, foreign_key: :job_uuid, primary_key: :uuid)
14 attr_accessor :job_readable
16 UNUSED_NODE_IP = '127.40.4.0'
18 api_accessible :user, :extend => :common do |t|
25 t.add :api_job_uuid, as: :job_uuid
26 t.add :crunch_worker_state
29 api_accessible :superuser, :extend => :user do |t|
32 t.add lambda { |x| Rails.configuration.compute_node_nameservers }, :as => :nameservers
36 super || Rails.configuration.compute_node_domain
40 job_readable ? job_uuid : nil
43 def crunch_worker_state
44 return 'down' if slot_number.nil?
45 case self.info.andand['slurm_state']
57 if db_current_time - self.created_at > 5.minutes
62 elsif db_current_time - self.last_ping_at > 1.hours
70 raise "must have :ip and :ping_secret" unless o[:ip] and o[:ping_secret]
72 if o[:ping_secret] != self.info['ping_secret']
73 logger.info "Ping: secret mismatch: received \"#{o[:ping_secret]}\" != \"#{self.info['ping_secret']}\""
74 raise ArvadosModel::UnauthorizedError.new("Incorrect ping_secret")
77 current_time = db_current_time
78 self.last_ping_at = current_time
80 @bypass_arvados_authorization = true
83 if self.ip_address.nil?
84 logger.info "#{self.uuid} ip_address= #{o[:ip]}"
85 self.ip_address = o[:ip]
86 self.first_ping_at = current_time
89 # Record instance ID if not already known
90 if o[:ec2_instance_id]
91 if !self.info['ec2_instance_id']
92 self.info['ec2_instance_id'] = o[:ec2_instance_id]
93 elsif self.info['ec2_instance_id'] != o[:ec2_instance_id]
94 logger.debug "Multiple nodes have credentials for #{self.uuid}"
95 raise "#{self.uuid} is already running at #{self.info['ec2_instance_id']} so rejecting ping from #{o[:ec2_instance_id]}"
100 if self.slot_number.nil?
103 self.slot_number = try_slot
107 rescue ActiveRecord::RecordNotUnique
110 raise "No available node slots" if try_slot == Rails.configuration.max_compute_nodes
115 if self.hostname.nil? and Rails.configuration.assign_node_hostname
116 self.hostname = self.class.hostname_for_slot(self.slot_number)
119 # Record other basic stats
120 ['total_cpu_cores', 'total_ram_mb', 'total_scratch_mb'].each do |key|
121 if value = (o[key] or o[key.to_sym])
122 self.properties[key] = value.to_i
124 self.properties.delete(key)
133 def ensure_ping_secret
134 self.info['ping_secret'] ||= rand(2**256).to_s(36)
137 def dns_server_update
138 if hostname_changed? && hostname_was
139 self.class.dns_server_update(hostname_was, UNUSED_NODE_IP)
141 if self.hostname_changed? or self.ip_address_changed?
142 if not self.ip_address.nil?
143 stale_conflicting_nodes = Node.where('id != ? and ip_address = ? and last_ping_at < ?',self.id,self.ip_address,10.minutes.ago)
144 if not stale_conflicting_nodes.empty?
145 # One or more stale compute node records have the same IP address as the new node.
146 # Clear the ip_address field on the stale nodes.
147 stale_conflicting_nodes.each do |stale_node|
148 stale_node.ip_address = nil
154 self.class.dns_server_update(hostname, ip_address || UNUSED_NODE_IP)
159 def self.dns_server_update hostname, ip_address
162 ptr_domain = ip_address.
163 split('.').reverse.join('.').concat('.in-addr.arpa')
167 uuid_prefix: Rails.configuration.uuid_prefix,
168 ip_address: ip_address,
169 ptr_domain: ptr_domain,
172 if Rails.configuration.dns_server_conf_dir and Rails.configuration.dns_server_conf_template
175 template = IO.read(Rails.configuration.dns_server_conf_template)
177 logger.error "Reading #{Rails.configuration.dns_server_conf_template}: #{e.message}"
181 hostfile = File.join Rails.configuration.dns_server_conf_dir, "#{hostname}.conf"
182 File.open hostfile+'.tmp', 'w' do |f|
183 f.puts template % template_vars
185 File.rename hostfile+'.tmp', hostfile
187 logger.error "Writing #{hostfile}: #{e.message}"
192 if Rails.configuration.dns_server_update_command
193 cmd = Rails.configuration.dns_server_update_command % template_vars
195 logger.error "dns_server_update_command #{cmd.inspect} failed: #{$?}"
200 if Rails.configuration.dns_server_conf_dir and Rails.configuration.dns_server_reload_command
201 restartfile = File.join(Rails.configuration.dns_server_conf_dir, 'restart.txt')
203 File.open(restartfile, 'w') do |f|
204 # Typically, this is used to trigger a dns server restart
205 f.puts Rails.configuration.dns_server_reload_command
208 logger.error "Unable to write #{restartfile}: #{e.message}"
216 def self.hostname_for_slot(slot_number)
217 config = Rails.configuration.assign_node_hostname
219 return nil if !config
221 sprintf(config, {:slot_number => slot_number})
224 # At startup, make sure all DNS entries exist. Otherwise, slurmctld
225 # will refuse to start.
226 if Rails.configuration.dns_server_conf_dir and Rails.configuration.dns_server_conf_template and Rails.configuration.assign_node_hostname
227 (0..Rails.configuration.max_compute_nodes-1).each do |slot_number|
228 hostname = hostname_for_slot(slot_number)
229 hostfile = File.join Rails.configuration.dns_server_conf_dir, "#{hostname}.conf"
230 if !File.exists? hostfile
231 n = Node.where(:slot_number => slot_number).first
232 if n.nil? or n.ip_address.nil?
233 dns_server_update(hostname, UNUSED_NODE_IP)
235 dns_server_update(hostname, n.ip_address)
241 def permission_to_update
242 @bypass_arvados_authorization or super
245 def permission_to_create
246 current_user and current_user.is_admin