site_hardware.rb 17.7 KB
Newer Older
1 2
# coding: utf-8

3
require 'refrepo/gpu_ref'
4

5 6 7 8 9 10
class SiteHardwareGenerator < WikiGenerator

  def initialize(page_name, site)
    super(page_name)
    @site = site
  end
11

12 13
  def generate_content
    @generated_content = "__NOTOC__\n__NOEDITSECTION__\n"
14
    @generated_content += "{{Portal|User}}\n"
15 16
    @generated_content += "<div class=\"sitelink\">[[Hardware|Global]] | " + G5K::SITES.map { |e| "[[#{e.capitalize}:Hardware|#{e.capitalize}]]" }.join(" | ") + "</div>\n"
    @generated_content += "\n= Summary =\n"
17
    @generated_content += "'''#{generate_oneline_summary}'''\n"
18 19
    @generated_content += self.class.generate_summary(@site, false)
    @generated_content += self.class.generate_description(@site)
20
    @generated_content += MW.italic(MW.small(generated_date_string))
21 22 23 24 25 26 27 28 29 30 31 32 33
    @generated_content += MW::LINE_FEED
  end

  def self.generate_all_clusters
    table_columns = []
    table_data = []
    G5K::SITES.each{ |site|
      table_columns = self.generate_summary_data(site, true)[0]
      table_data += self.generate_summary_data(site, true)[1]
    }
    MW.generate_table('class="wikitable sortable"', table_columns, table_data) + "\n"
  end

34 35 36 37 38
  def generate_oneline_summary
    h = G5K::get_global_hash['sites'][@site]
    # remove retired nodes
    # FIXME this should probably move to a helper
    h['clusters'].each_pair do |cl, v|
Lucas Nussbaum's avatar
Lucas Nussbaum committed
39
      v['nodes'].delete_if { |n, v2| v2['status'] == 'retired' }
40 41 42 43 44 45 46 47 48 49 50
    end
    h['clusters'].delete_if { |k, v| v['nodes'].empty? }

    clusters = h['clusters'].length
    nodes = h['clusters'].inject(0) { |a, b| a + b[1]['nodes'].values.length }
    cores = h['clusters'].inject(0) { |a, b| cnodes = b[1]['nodes'].values ; a + cnodes.length * cnodes.first['architecture']['nb_cores'] }
    flops = h['clusters'].inject(0) { |a, b| cnodes = b[1]['nodes'].values ; a + cnodes.length * (cnodes.first['performance']['node_flops'] rescue 0) }
    tflops = sprintf("%.1f", flops.to_f / (10**12))
    return "#{clusters} cluster#{clusters > 1 ? 's' : ''}, #{nodes} node#{nodes > 1 ? 's' : ''}, #{cores} core#{cores > 1 ? 's' : ''}, #{tflops} TFLOPS"
  end

51 52 53 54
  def self.generate_summary(site, with_sites)
    table_columns, table_data = self.generate_summary_data(site, with_sites)
    MW.generate_table('class="wikitable sortable"', table_columns, table_data) + "\n"
  end
55

56 57 58 59 60 61 62 63 64 65
  def self.generate_summary_data(site, with_sites)
    table_columns = []
    table_data = []

    hardware = get_hardware([site])

    site_accelerators = 0
    hardware[site].sort.to_h.each { |cluster_uid, cluster_hash|
      site_accelerators += cluster_hash.select { |k, v| v['accelerators'] != '' }.count
    }
66

67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86
    hardware[site].sort.to_h.each { |cluster_uid, cluster_hash|
      cluster_nodes = cluster_hash.keys.flatten.count
      queue = cluster_hash.map { |k, v| v['queue']}.first
      queue_str = cluster_hash.map { |k, v| v['queue_str']}.first
      table_columns = (with_sites == true ? ['Site'] : []) + ['Cluster',  'Queue', 'Date of arrival', { attributes: 'data-sort-type="number"', text: 'Nodes' }, 'CPU', { attributes: 'data-sort-type="number"', text: 'Cores' }, { attributes: 'data-sort-type="number"', text: 'Memory' }, { attributes: 'data-sort-type="number"', text: 'Storage' }, { attributes: 'data-sort-type="number"', text: 'Network' }] + ((site_accelerators.zero? && with_sites == false) ? [] : ['Accelerators'])
      data = partition(cluster_hash)
      table_data <<  (with_sites == true ? ["[[#{site.capitalize}:Hardware|#{site.capitalize}]]"] : []) + [
        (with_sites == true ? "[[#{site.capitalize}:Hardware##{cluster_uid}" + (queue_str == '' ? '' : "_.28#{queue_str.gsub(' ', '_')}.29") + "|#{cluster_uid}]]" : "[[##{cluster_uid}" + (queue_str == '' ? '' : "_.28#{queue_str.gsub(' ', '_')}.29") + "|#{cluster_uid}]]"),
        (queue == '' ? 'default' : queue),
        cell_data(data, 'date'),
        cluster_nodes,
        cell_data(data, 'num_processor_model'),
        cell_data(data, 'cores_per_cpu_str'),
        cell_data(data, 'ram_size'),
        'data-sort-value="' + sort_data(data, 'storage_size') + '"|' + cell_data(data, 'storage'),
        'data-sort-value="' + sort_data(data, 'network_throughput') + '"|' + cell_data(data, 'used_networks')
      ] + ((site_accelerators.zero? && with_sites == false) ? [] : [cell_data(data, 'accelerators')])
    }
    [table_columns, table_data]
  end
87

88 89 90 91 92 93 94 95 96 97
  def self.generate_description(site)
    table_columns = []
    text_data = []

    hardware = get_hardware([site])

    site_accelerators = 0
    hardware[site].sort.to_h.each { |cluster_uid, cluster_hash|
      site_accelerators += cluster_hash.select { |k, v| v['accelerators'] != '' }.count
    }
98

99 100 101 102 103 104 105
    hardware[site].sort.to_h.each { |cluster_uid, cluster_hash|
      subclusters = cluster_hash.keys.count != 1
      cluster_nodes = cluster_hash.keys.flatten.count
      cluster_cpus = cluster_hash.map { |k, v| k.count * v['cpus_per_node'] }.reduce(:+)
      cluster_cores = cluster_hash.map { |k, v| k.count * v['cpus_per_node'] * v['cores_per_cpu'] }.reduce(:+)
      queue_str = cluster_hash.map { |k, v| v['queue_str']}.first
      table_columns = ['Cluster',  'Queue', 'Date of arrival', { attributes: 'data-sort-type="number"', text: 'Nodes' }, 'CPU', { attributes: 'data-sort-type="number"', text: 'Cores' }, { attributes: 'data-sort-type="number"', text: 'Memory' }, { attributes: 'data-sort-type="number"', text: 'Storage' }, { attributes: 'data-sort-type="number"', text: 'Network' }] + (site_accelerators.zero? ? [] : ['Accelerators'])
106

107
      text_data <<  ["\n== #{cluster_uid}" + (queue_str == '' ? '' : " (#{queue_str})") + " ==\n"]
108
      text_data << ["'''#{cluster_nodes} #{G5K.pluralize(cluster_nodes, 'node')}, #{cluster_cpus} #{G5K.pluralize(cluster_cpus, 'cpu')}, #{cluster_cores} #{G5K.pluralize(cluster_cores, 'core')}" + (subclusters == true ? ",''' split as follows due to differences between nodes " : "''' ") + "([https://public-api.grid5000.fr/stable/sites/#{site}/clusters/#{cluster_uid}/nodes.json?pretty=1 json])"]
109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141

      cluster_hash.sort.to_h.each_with_index { |(num, h), i|
        if subclusters
          subcluster_nodes = num.count
          subcluster_cpus = subcluster_nodes * h['cpus_per_node']
          subcluster_cores = subcluster_nodes * h['cpus_per_node'] * h['cores_per_cpu']
          text_data << "<hr style=\"height:10pt; visibility:hidden;\" />\n" if i != 0 # smaller vertical <br />
          text_data << ["; #{cluster_uid}-#{G5K.nodeset(num)} (#{subcluster_nodes} #{G5K.pluralize(subcluster_nodes, 'node')}, #{subcluster_cpus} #{G5K.pluralize(subcluster_cpus, 'cpu')}, #{subcluster_cores} #{G5K.pluralize(subcluster_cores, 'core')})"]
        end

        accelerators = nil
        if h['gpu_str'] != '' && h['mic_str'] != ''
          accelerators = 'GPU/Xeon Phi'
        elsif h['gpu_str'] != ''
          accelerators = 'GPU'
        elsif h['mic_str'] != ''
          accelerators = 'Xeon Phi'
        end
        hash = {
          'Model' => h['model'],
          'Date of arrival' => h['date'],
          'CPU' => h['processor_description'],
          'Memory' => h['ram_size'],
          'Storage' => h['storage_description'],
          'Network' => h['network_description'],
        }
        hash[accelerators] = h['accelerators'] if accelerators
        text_data << MW::generate_hash_table(hash)
      }
    }

    generated_content = "\n= Cluster details =\n"
    generated_content += text_data.join("\n")
Lucas Nussbaum's avatar
Lucas Nussbaum committed
142
    return generated_content
143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177
  end
end

# Partitioning the hash values
def partition(cluster_hash)
  data = {}
  h1 = {}
  cluster_hash.sort.to_h.each { |num2, h2|
    h2.each_key{ |k|
      h1[k] = []
      cluster_hash.sort.to_h.each { |num3, h3|
        if h1[k].map {|e| e['value']}.include?(h3[k])
          num = (h1[k].select { |e| e['value'] == h3[k] }).first['num']
          h1[k].delete_if { |e| e['value'] == h3[k] }
          h1[k] += [{ 'num' => num + num3, 'value' => h3[k], 'cell' => "#{G5K.nodeset(num + num3)}: #{h3[k]}", 'sort' => "#{h3[k]}"}]
        else
          h1[k] += [{ 'num' => num3, 'value' => h3[k], 'cell' => "#{G5K.nodeset(num3)}: #{h3[k]}", 'sort' => "#{h3[k]}"}]
        end
      }
      h1[k].first['cell'] = h1[k].first['cell'].split(': ')[1] if h1[k].count == 1
      data[k] = h1[k].sort_by{ |e| e['num'].sort[0] }
    }
  }
  data
end

def cell_data(data, key)
  data[key].map{ |e| e['cell'] }.join('<br />')
end

def sort_data(data, key)
  data[key].map{ |e| e['sort'] }[0]
end

def get_hardware(sites)
178
  global_hash = G5K::get_global_hash
179

180 181 182 183 184 185 186 187
  # Loop over each cluster of the site
  hardware = {}
  global_hash['sites'].sort.to_h.select{ |site_uid, site_hash| sites.include?(site_uid) }.each { |site_uid, site_hash|
    hardware[site_uid] = {}
    site_hash['clusters'].sort.to_h.each { |cluster_uid, cluster_hash|
      hardware[site_uid][cluster_uid] = {}
      cluster_hash.fetch('nodes').sort.each { |node_uid, node_hash|
        next if node_hash['status'] == 'retired'
188
        # map model to vendor (eg: {'SAS5484654' => 'Seagate', 'PX458' => 'Toshiba' ...}
189 190 191 192 193 194 195 196
        hard = {}
        queue = cluster_hash['queues'] - ['admin', 'default']
        hard['queue'] = (queue.nil? || queue.empty?) ? '' : queue[0]
        hard['queue_str'] = (queue.nil? || queue.empty?) ? '' : queue[0] + G5K.pluralize(queue.count, ' queue')

        hard['date'] = Date.parse(cluster_hash['created_at'].to_s).strftime('%Y-%m-%d')
        hard['model'] = cluster_hash['model']
        hard['processor_model'] = [node_hash['processor']['model'], node_hash['processor']['version']].join(' ')
197 198 199
        if node_hash['processor']['other_description'] =~ /@/
          hard['processor_freq'] = node_hash['processor']['other_description'].split('@')[1].strip
        end
200 201 202 203 204 205
        hard['microarchitecture'] = node_hash['processor']['microarchitecture']
        hard['cpus_per_node'] = node_hash['architecture']['nb_procs']
        hard['cpus_per_node_str'] = hard['cpus_per_node'].to_s + '&nbsp;' + G5K.pluralize(hard['cpus_per_node'], 'CPU') + '/node'
        hard['cores_per_cpu'] = node_hash['architecture']['nb_cores'] / hard['cpus_per_node']
        hard['cores_per_cpu_str'] = hard['cores_per_cpu'].to_s + '&nbsp;' + G5K.pluralize(hard['cores_per_cpu'], 'core') + '/CPU'
        hard['num_processor_model'] = (hard['cpus_per_node'] == 1 ? '' : "#{hard['cpus_per_node']}&nbsp;x&nbsp;") + hard['processor_model'].gsub(' ', '&nbsp;')
206
        hard['processor_description'] = "#{hard['processor_model']} (#{hard['microarchitecture']}#{hard['processor_freq'] ?  ', ' + hard['processor_freq'] : ''}, #{hard['cpus_per_node_str']}, #{hard['cores_per_cpu_str']})"
207
        hard['ram_size'] = G5K.get_size(node_hash['main_memory']['ram_size'])
208 209 210 211 212 213 214 215 216 217 218 219
        storage = node_hash['storage_devices'].map { |i| { 'size' => i['size'], 'tech' => i['storage'] } }
        hard['storage'] = storage.each_with_object(Hash.new(0)) { |data, counts|
          counts[data] += 1
        }.to_a.sort_by { |e|
          e[0]['size'].to_f
        }.map { |e|
          (e[1] == 1 ? '' : e[1].to_s + '&nbsp;x&nbsp;') +
          G5K.get_size(e[0]['size'], 'metric') +
          '&nbsp;' +
          e[0]['tech']
        }.join(' +&nbsp;')

220
        hard['storage_size'] = storage.inject(0){|sum, v| sum + (v['size'].to_f / 2**30).floor }.to_s # round to GB to avoid small differences within a cluster
221 222 223 224 225 226 227 228 229 230 231 232 233 234 235
        storage_description = node_hash['storage_devices'].sort { |a,b|
          a['device'] <=> b['device']
        }.map do |v|
          {
            'device' => v['device'],
            'size' => v['size'],
            'tech' => v['storage'],
            'interface' => v['interface'],
            'vendor' => v['vendor'],
            'model' => v['model'],
            'driver' => v['driver'],
            'path' => v['by_path'] || v['by_id'],
            'count' => node_hash['storage_devices'].count
          }
        end
236

237 238 239 240 241 242 243 244 245 246 247
        hard['storage_description'] = storage_description.map { |e|
          [
            e['count'] > 1 ? "\n*" : '',
            G5K.get_size(e['size'],'metric'),
            e['tech'],
            e['interface'],
            e['vendor'],
            e['model'],
            ' (driver: ' + (e['driver'] || 'MISSING') + ', path: ' + (e['path'] || 'MISSING') + ')'
          ].join(' ')
        }.join('<br />')
248

249
        network = node_hash['network_adapters'].select { |v|
250
          v['management'] == false &&
251 252
          (v['device'] =~ /\./).nil? # exclude PKEY / VLAN interfaces see #9417
        }.map{|v| {
253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273
            'rate' => v['rate'],
            'interface' => v['interface'],
            'used' => (v['enabled'] and (v['mounted'] or v['mountable']))
          }
        }
        hard['used_networks'] = network.select { |e|
          e['used'] == true
        }.each_with_object(Hash.new(0)) { |data, counts|
          counts[data] += 1
        }.to_a.sort_by { |e|
          e[0]['rate'].to_f
        }.map{ |e|
          get_network_info(e, false)
        }.join('+&nbsp;')

        hard['network_throughput'] = network.select { |e|
          e['used'] == true
        }.inject(0){ |sum, v|
          sum + (v['rate'].to_f / 10**6).floor
        }.to_s # round to Mbps

274
        network_description = node_hash['network_adapters'].select { |v|
275
          v['management'] == false &&
276 277
          (v['device'] =~ /\./).nil? # exclude PKEY / VLAN interface see #9417
        }.map{ |v|
278
          {
279
            'device' => v['device'],
280 281 282 283 284 285 286 287 288 289 290 291 292 293 294
            'name' => v['name'],
            'rate' => v['rate'],
            'interface' => v['interface'],
            'driver' => v['driver'],
            'unwired' => v['enabled'] == false,
            'unavailable_for_experiment' => v['mountable'] == false,
            'no_kavlan' => (v['interface'] == 'Ethernet' &&
              v['mountable'] == true &&
              v['kavlan'] == false),
            'model' => (v['vendor'] || 'N/A') + ' ' + (v['model'] || 'N/A'),
            'count' => node_hash['network_adapters'].count
          }
        }.sort_by{ |e|
          e['device']
        }
295 296 297
        hard['network_description'] = network_description.map do |e|
          s  = e['count'] > 1 ? "\n* " : ''
          s += e['unavailable_for_experiment'] ? '<span style="color:grey">' : ''
298 299 300 301 302
          if e['name'].nil? or e['name'] == e['device']
            s += e['device']
          else
            s +=  e['device'] + "/" + e['name']
          end
303 304 305 306 307 308 309 310 311 312 313
          s += ', '
          s += e['interface']
          s += ', '
          if !(e['unwired'] and e['unavailable_for_experiment'])
            s += 'configured rate: ' + (e['unwired'] ? 'n/c' : G5K.get_rate(e['rate']))
            s += ', '
          end
          if !(e['model'] == 'N/A N/A' and e['unavailable_for_experiment']) # don't include interface model if not available
            e['model'] = 'N/A' if e['model'] == 'N/A N/A'
            s += 'model: '+ e['model'] + ', '
          end
314
          s +=  'driver: ' + e['driver'] if e['driver']
315 316 317 318 319
          s += ' - unavailable for experiment' if e['unavailable_for_experiment']
          s += ' - no KaVLAN' if e['no_kavlan']
          s +=  e['unavailable_for_experiment'] ? '</span>' : ''
          s
        end.join('<br />')
320

321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339
        lgpu = node_hash['gpu_devices']
        hard['gpu_str'] = if lgpu
                            bymodel = {}
                            lgpu.each { |g|
                              d = g[1]
                              vendor = d['vendor']
                              model = GPURef.getGrid5000LegacyNameFor(d['model'])
                              vm = vendor.to_s + ' ' + model.to_s.gsub(' ', '&nbsp;')
                              if bymodel[vm]
                                bymodel[vm] += 1
                              else
                                bymodel[vm] = 1
                              end
                            }
                            res = []
                            bymodel.each { |model,count|
                              res << (count == 1 ? '' : count.to_s + '&nbsp;x&nbsp;') + model
                            }
                            res.join(", ")
340 341 342 343 344 345 346 347 348 349
                          else
                            ''
                          end
        mic = node_hash['mic']
        hard['mic_str'] = if mic
                            (mic['mic_count'].to_i == 1 ? '' : mic['mic_count'].to_s + '&nbsp;x&nbsp;') + mic['mic_vendor'].to_s + ' ' + mic['mic_model'].to_s.gsub(' ', '&nbsp;')
                          else
                            ''
                          end
        hard['accelerators'] = hard['gpu_str'] != '' ? hard['gpu_str'] + (hard['mic_str'] != '' ? ' ; ' + hard['mic_str'] : '') : hard['mic_str']
350

351
        add(hardware[site_uid][cluster_uid], node_uid, hard)
352
      }
353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369 370 371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389
    }
  }
  hardware
end

# This methods adds the array hard to the hash
# hardware. If nodes 2,3,7 have the same hard, they
# will be gathered in the same key and we will have
# hardware[[2,3,7]] = hard
def add(hardware, node_uid, hard)
  num1 = node_uid.split('-')[1].to_i
  if hardware.has_value?(hard) == false
    hardware[[num1]] = hard
  else
    num2 = hardware.key(hard)
    hardware.delete(num2)
    hardware[num2.push(num1)] = hard
  end
end

def get_network_info(e, all_networks)
  rate = G5K.get_rate(e[0]['rate'])
  (e[1] == 1 ? '' : e[1].to_s + '&nbsp;x&nbsp;') +
    (rate == '' ? '' : rate + '&nbsp;') +
    (all_networks ? e[0]['interface'].to_s : get_interface(e[0]['interface'])) +
    (all_networks ? (e[0]['used'] == true ? '' : ' (unused)') : '')
end

def get_interface(interface)
  return interface == 'Ethernet' ? '' : interface
end

# Only execute if the file is run directly
if __FILE__ == $0
  options = WikiGenerator::parse_options

  if (options)
390 391 392 393 394
    ret=2
    begin
      ret = true
      generators = options[:sites].map{ |site| SiteHardwareGenerator.new(site.capitalize + ':Hardware', site) }
      generators.each{ |generator|
395
        ret &= generator.exec(options)
396 397 398 399 400 401 402 403 404 405
      }
    rescue MediawikiApi::ApiError => e
      puts e, e.backtrace
      ret = 3
    rescue StandardError => e
      puts e, e.backtrace
      ret = 4
    ensure
      exit(ret)
    end
406 407
  end
end