Attention une mise à jour du serveur va être effectuée le lundi 17 mai entre 13h et 13h30. Cette mise à jour va générer une interruption du service de quelques minutes.

hardware.rb 17.2 KB
Newer Older
1
# coding: utf-8
2
$LOAD_PATH.unshift(File.expand_path(File.join(File.dirname(__FILE__), 'lib')))
3
require 'refrepo/gen/wiki/wiki_generator'
4 5 6 7 8 9 10 11 12
require_relative './site_hardware.rb'

class G5KHardwareGenerator < WikiGenerator

  def initialize(page_name)
    super(page_name)
  end

  def generate_content
13
    @global_hash = get_global_hash
14
    @site_uids = G5K::SITES
15

16
    @generated_content = "__NOEDITSECTION__\n"
17
    @generated_content += "<div class=\"sitelink\">[[Hardware|Global]] | " + G5K::SITES.map { |e| "[[#{e.capitalize}:Hardware|#{e.capitalize}]]" }.join(" | ") + "</div>\n"
18 19 20
    @generated_content += "\n= Clusters =\n"
    @generated_content += SiteHardwareGenerator.generate_all_clusters
    @generated_content += generate_totals
21
    @generated_content += MW.italic(MW.small(generated_date_string))
22 23 24 25 26 27 28 29 30 31
    @generated_content += MW::LINE_FEED
  end

  def generate_totals
    data = {
      'proc_families' => {},
      'proc_models' => {},
      'core_models' => {},
      'ram_size' => {},
      'net_interconnects' => {},
32
      'net_models' => {},
33 34 35 36 37
      'acc_families' => {},
      'acc_models' => {},
      'acc_cores' => {},
      'node_models' => {}
    }
38

39 40 41
    @global_hash['sites'].sort.to_h.each { |site_uid, site_hash|
      site_hash['clusters'].sort.to_h.each { |cluster_uid, cluster_hash|
        cluster_hash['nodes'].sort.to_h.each { |node_uid, node_hash|
42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150
          begin
            next if node_hash['status'] == 'retired'
            @node = node_uid

            # Processors
            model = node_hash['processor']['model']
            version = "#{model} #{node_hash['processor']['version']}"
            microarchitecture = node_hash['processor']['microarchitecture']

            cluster_procs = node_hash['architecture']['nb_procs']
            cluster_cores = node_hash['architecture']['nb_cores']

            key = [model]
            init(data, 'proc_families', key)
            data['proc_families'][key][site_uid] += cluster_procs

            key = [{text: microarchitecture || ' ', sort: get_date(microarchitecture) + ', ' + microarchitecture.to_s}, {text: version, sort: get_date(microarchitecture) + ', ' + version.to_s}]
            init(data, 'proc_models', key)
            data['proc_models'][key][site_uid] += cluster_procs

            init(data, 'core_models', key)
            data['core_models'][key][site_uid] += cluster_cores

            # RAM size
            ram_size = node_hash['main_memory']['ram_size']
            key = [{ text: G5K.get_size(ram_size), sort: (ram_size / 2**30).to_s.rjust(6, '0') + ' GB' }]
            init(data, 'ram_size', key)
            data['ram_size'][key][site_uid] += 1

            # HPC Networks
            interfaces = node_hash['network_adapters'].select{ |k, v|
              v['enabled'] and
                (v['mounted'] or v['mountable']) and
                not v['management'] and
                (k =~ /\./).nil? # exclude PKEY / VLAN interfaces see #9417
            }.map{ |k, v|
              [
                {
                  text: v['interface'] + ' ' + G5K.get_rate(v['rate']),
                  sort: ((v['rate'])/10**6).to_s.rjust(6, '0') + ' Gbps, ' + v['interface']
                }
              ]
            }.uniq

            net_interconnects = interfaces.inject(Hash.new(0)){ |h, v| h[v] += 1; h }
            net_interconnects.sort_by { |k, v|  k.first[:sort] }.each { |k, v|
              init(data, 'net_interconnects', k)
              data['net_interconnects'][k][site_uid] += v
            }

            # NIC models
            interfaces = node_hash['network_adapters'].select{ |k, v|
              v['enabled'] and
                (v['mounted'] or v['mountable']) and
                not v['management'] and
                (k =~ /\./).nil? # exclude PKEY / VLAN interfaces see #9417
            }.map{ |k, v|
              t = (v['vendor'] || 'N/A') + ' ' + (v['model'] || 'N/A');
              [
                {
                  text: v['interface'],
                  sort: v['interface']
                },
                {
                  text: t, sort: t
                }
              ]
            }.uniq

            net_models = interfaces.inject(Hash.new(0)){ |h, v| h[v] += 1; h }
            net_models.sort_by { |k, v|  k.first[:sort] }.each { |k, v|
              init(data, 'net_models', k)
              data['net_models'][k][site_uid] += v
            }

            # Accelerators
            g = node_hash['gpu']
            m = node_hash['mic']

            gpu_families = {}
            gpu_families[[g['gpu_vendor']]] = g['gpu_count'] if g and g['gpu']
            mic_families = {}
            mic_families[[m['mic_vendor']]] = m['mic_count'] if m and m['mic']
            gpu_families.merge(mic_families).sort.to_h.each { |k, v|
              init(data, 'acc_families', k)
              data['acc_families'][k][site_uid] += v
            }

            gpu_details = {}
            gpu_details[["#{g['gpu_vendor']} #{g['gpu_model']}"]] = [g['gpu_count'], g['gpu_cores']] if g and g['gpu']
            mic_details = {}
            mic_details[["#{m['mic_vendor']} #{m['mic_model']}"]] = [m['mic_count'], m['mic_cores']] if m and m['mic']

            gpu_details.merge(mic_details).sort.to_h.each { |k, v|
              init(data, 'acc_models', k)
              data['acc_models'][k][site_uid] += v[0]

              init(data, 'acc_cores', k)
              data['acc_cores'][k][site_uid] += v[1]
            }

            # Nodes
            key = [cluster_hash['model']]
            init(data, 'node_models', key)
            data['node_models'][key][site_uid] += 1
          rescue
            puts "ERROR while processing #{node_uid}: #{$!}"
            raise
          end
151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172
        }
      }
    }

    # Table construction
    generated_content = "= Processors ="
    generated_content += "\n== Processors counts per families ==\n"
    sites = @site_uids.map{ |e| "[[#{e.capitalize}:Hardware|#{e.capitalize}]]" }
    table_options = 'class="wikitable sortable" style="text-align: center;"'
    table_columns = ['Processor family'] + sites + ['Processors total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'proc_families'))
    generated_content += "\n== Processors counts per models ==\n"
    table_columns = ['Microarchitecture', 'Processor model'] + sites + ['Processors total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'proc_models'))
    generated_content += "\n== Cores counts per models ==\n"
    table_columns =  ['Microarchitecture', 'Core model'] + sites + ['Cores total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'core_models'))

    generated_content += "\n= RAM size per node =\n"
    table_columns = ['RAM size'] + sites + ['Nodes total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'ram_size'))

173 174
    generated_content += "\n= Networking =\n"
    generated_content += "\n== Network interconnects ==\n"
175 176 177
    table_columns = ['Interconnect'] + sites + ['Cards total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'net_interconnects'))

178
    generated_content += "\n== Nodes with several Ethernet interfaces ==\n"
179
    generated_content +=  generate_interfaces
180 181

    generated_content += "\n== Network interface models ==\n"
182
    table_columns = ['Type', 'Model'] + sites + ['Cards total']
183
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'net_models'))
184

185 186 187 188
    generated_content += "\n= Storage ="
    generated_content += "\n== Nodes with several disks ==\n"
    generated_content +=  generate_storage

189 190 191 192 193 194 195 196 197 198
    generated_content += "\n= Accelerators (GPU, Xeon Phi) ="
    generated_content += "\n== Accelerator families ==\n"
    table_columns = ['Accelerator family'] + sites + ['Accelerators total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'acc_families'))
    table_columns = ['Accelerator model'] + sites + ['Accelerators total']
    generated_content += "\n== Accelerator models ==\n"
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'acc_models'))
    generated_content += "\n== Accelerator cores ==\n"
    table_columns = ['Accelerator model'] + sites + ['Cores total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'acc_cores'))
199

200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219
    generated_content += "\n= Nodes models =\n"
    table_columns = ['Nodes model'] + sites + ['Nodes total']
    generated_content += MW.generate_table(table_options, table_columns, get_table_data(data, 'node_models'))
  end

  def init(data, key1, key2)
    if not data[key1].key?(key2)
      data[key1][key2] = {}
      @site_uids.each { |s| data[key1][key2][s] = 0 }
    end
  end

  # This method generates a wiki table from data[key] values, sorted by key
  # values in first column.
  def get_table_data(data, key)
    raw_data = []
    table_data = []
    index = 0
    k0 = 0
    data[key].sort_by{
220 221 222 223
      # Sort the table by the identifiers (e.g. Microarchitecture, or Microarchitecture + CPU name).
      # This colum is either just a text field, or a more complex hash with a :sort key that should be
      # used for sorting.
      |k, v| k.map { |c| c.kind_of?(Hash) ? c[:sort] : c }
224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243
    }.to_h.each { |k, v|
      k0 = k if index == 0
      index += 1
      elts = v.sort.to_h.values
      raw_data << elts
      table_data << k.map{ |e| e.kind_of?(Hash) ? "data-sort-value=\"#{e[:sort]}\"|#{e[:text]}" : "data-sort-value=\"#{index.to_s.rjust(3, '0')}\"|#{e}" } +
        elts.map{ |e| e.kind_of?(Hash) ? "data-sort-value=\"#{e[:sort]}\"|#{e[:text]}" : e }
        .map{ |e| e == 0 ? '' : e  } + ["'''#{elts.reduce(:+)}'''"]
    }
    elts = raw_data.transpose.map{ |e| e.reduce(:+)}
    table_data << {columns: ["'''Sites total'''"] +
                   [' '] * (k0.length - 1) +
                   (elts + [elts.reduce(:+)]).map{ |e| e == 0 ? '' : "'''#{e}'''" },
                   sort: false}
  end

  # See: https://en.wikipedia.org/wiki/List_of_Intel_Xeon_microprocessors
  # For a correct sort of the column, all dates must be in the same
  # format (same number of digits)
  def get_date(microarchitecture)
244
    return 'MISSING' if microarchitecture.nil?
245 246 247 248 249 250 251 252 253 254 255 256
    release_dates = {
      'K8' => '2003',
      'K10' => '2007',
      'Clovertown' => '2006',
      'Harpertown' => '2007',
      'Dunnington' => '2008',
      'Lynnfield' => '2009',
      'Nehalem' => '2010',
      'Westmere' => '2011',
      'Sandy Bridge' => '2012',
      'Haswell' => '2013',
      'Broadwell' => '2015',
257 258
      'Skylake' => '2016',
      'Zen' => '2017'
259 260
    }
    date = release_dates[microarchitecture.to_s]
261
    raise "ERROR: microarchitecture not found: '#{microarchitecture.to_s}'. Add in hardware.rb" if date.nil?
262 263
    date
  end
264 265 266 267 268 269 270 271 272 273

  def storage_size_to_text(s)
    if s > 1000*1000*1000*1000 # 1 TB
      return sprintf("%.1f", s/(1000*1000*1000*1000)) + ' TB'
    else
      return sprintf("%d", s/(1000*1000*1000)) + ' GB'
    end
  end

  def generate_storage
274
    table_columns = ["Site", "Cluster", "Number of nodes", "Main disk", "Additional HDDs", "Additional SSDs", "[[Disk_reservation|Disk reservation]]"]
275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300
    table_data = []
    global_hash = get_global_hash

    # Loop over Grid'5000 sites
    global_hash["sites"].sort.to_h.each do |site_uid, site_hash|
      site_hash.fetch("clusters").sort.to_h.each do |cluster_uid, cluster_hash|
        nodes_data = []
        cluster_hash.fetch('nodes').sort.to_h.each do |node_uid, node_hash|
          next if node_hash['status'] == 'retired'
          sd = node_hash['storage_devices']
          reservable_disks = sd.to_a.select{ |v| v[1]['reservation'] == true }.count > 0
          maindisk = sd.to_a.select { |v| v[0] == 'sda' }.first[1]
          maindisk_t = maindisk['storage'] + ' ' + storage_size_to_text(maindisk['size'])
          other = sd.to_a.select { |d| d[0] != 'sda' }
          hdds = other.select { |d| d[1]['storage'] == 'HDD' }
          if hdds.count == 0
            hdd_t = "0"
          else
            hdd_t = hdds.count.to_s + " (" + hdds.map { |d| storage_size_to_text(d[1]['size']) }.join(', ') + ")"
          end
          ssds = other.select { |d| d[1]['storage'] == 'SSD' }
          if ssds.count == 0
            ssd_t = "0"
          else
            ssd_t = ssds.count.to_s + " (" + ssds.map { |d| storage_size_to_text(d[1]['size']) }.join(', ') + ")"
          end
301 302 303
          queues = cluster_hash['queues'] - ['admin', 'default']
          queue_t = (queues.nil? || (queues.empty? ? '' : "_.28" + queues[0].gsub(' ', '_') + ' queue.29'))
          nodes_data << { 'uid' => node_uid, 'data' => { 'main' => maindisk_t, 'hdd' => hdd_t, 'ssd' => ssd_t, 'reservation' => reservable_disks, 'queue' => queue_t } }
304 305 306 307 308 309 310 311 312 313 314 315
        end
        nd = nodes_data.group_by { |d| d['data'] }
        nd.each do |data, nodes|
          # only keep nodes with more than one disk
          next if data['hdd'] == "0" and data['ssd'] == "0"
          if nd.length == 1
            nodesetname = cluster_uid
          else
            nodesetname = G5K.nodeset(nodes.map { |n| n['uid'].split('-')[1].to_i })
          end
          table_data << [
            "[[#{site_uid.capitalize}:Hardware|#{site_uid.capitalize}]]",
316
              "[[#{site_uid.capitalize}:Hardware##{cluster_uid}#{data['queue']}|#{nodesetname}]]",
317 318 319 320 321 322 323 324 325 326 327 328 329 330 331
              nodes.length,
              data['main'],
              data['hdd'],
              data['ssd'],
              data['reservation'] ? 'yes' : 'no'
          ]
        end
      end
    end
    # Sort by site and cluster name
    table_data.sort_by! { |row|
      [row[0], row[1]]
    }

    # Table construction
Lucas Nussbaum's avatar
Lucas Nussbaum committed
332
    table_options = 'class="wikitable sortable" style="text-align: center;"'
333 334 335
    return MW.generate_table(table_options, table_columns, table_data)
  end

336 337
  def generate_interfaces
    table_data = []
338 339
    @global_hash["sites"].sort.to_h.each { |site_uid, site_hash|
      site_hash.fetch("clusters").sort.to_h.each { |cluster_uid, cluster_hash|
340
        network_interfaces = {}
341
        cluster_hash.fetch('nodes').sort.to_h.each { |node_uid, node_hash|
342 343 344 345 346
          next if node_hash['status'] == 'retired'
          if node_hash['network_adapters']
            node_interfaces = node_hash['network_adapters'].select{ |k, v| v['interface'] == 'Ethernet' and v['enabled'] == true and (v['mounted'] == true or v['mountable'] == true) and v['management'] == false }

            interfaces = {}
DELABROYE Dimitri's avatar
DELABROYE Dimitri committed
347
            interfaces['25g_count'] = node_interfaces.select { |k, v| v['rate'] == 25_000_000_000 }.count
348 349 350
            interfaces['10g_count'] = node_interfaces.select { |k, v| v['rate'] == 10_000_000_000 }.count
            interfaces['1g_count'] = node_interfaces.select { |k, v| v['rate'] == 1_000_000_000 }.count
            interfaces['details'] = node_interfaces.map{ |k, v| k + (v['name'].nil? ? '' : '/' + v['name'])  + ' (' + G5K.get_rate(v['rate']) + ')' }.sort.join(', ')
351
            queues = cluster_hash['queues'] - ['admin', 'default', 'testing']
352
            interfaces['queues'] = (queues.nil? || (queues.empty? ? '' : queues[0] + G5K.pluralize(queues.count, ' queue')))
353
            interface_add(network_interfaces, node_uid, interfaces) if node_interfaces.count > 1
354 355 356 357 358 359 360 361 362
          end
        }

        # One line for each group of nodes with the same interfaces
        network_interfaces.sort.to_h.each { |num, interfaces|
          table_data << [
            "[[#{site_uid.capitalize}:Network|#{site_uid.capitalize}]]",
            "[[#{site_uid.capitalize}:Hardware##{cluster_uid}" + (interfaces['queues'] == '' ? '' : "_.28#{queues.gsub(' ', '_')}.29") + "|#{cluster_uid}" + (network_interfaces.size==1 ? '' : '-' + G5K.nodeset(num)) + "]]",
            num.count,
DELABROYE Dimitri's avatar
DELABROYE Dimitri committed
363
            interfaces['25g_count'].zero? ? '' : interfaces['25g_count'],
364 365 366 367 368 369 370 371 372 373 374 375 376
            interfaces['10g_count'].zero? ? '' : interfaces['10g_count'],
            interfaces['1g_count'].zero? ? '' : interfaces['1g_count'],
            interfaces['details']
          ]
        }
      }
    }
    # Sort by site and cluster name
    table_data.sort_by! { |row|
      [row[0], row[1]]
    }

    table_options = 'class="wikitable sortable" style="text-align: center;"'
377
    table_columns = ["Site", "Cluster", "Nodes", "25G interfaces", "10G interfaces", "1G interfaces", "Interfaces (throughput)"]
378
    MW.generate_table(table_options, table_columns, table_data)
379 380 381 382 383 384
  end

  # This methods adds the array interfaces to the hash
  # network_interfaces. If nodes 2,3,7 have the same interfaces, they
  # will be gathered in the same key and we will have
  # network_interfaces[[2,3,7]] = interfaces
385
  def interface_add(network_interfaces, node_uid, interfaces)
386 387 388 389 390 391 392 393
    num1 = node_uid.split('-')[1].to_i
    if network_interfaces.has_value?(interfaces) == false
      network_interfaces[[num1]] = interfaces
    else
      num2 = network_interfaces.key(interfaces)
      network_interfaces.delete(num2)
      network_interfaces[num2.push(num1)] = interfaces
    end
394
  end
395 396
end

Lucas Nussbaum's avatar
Lucas Nussbaum committed
397 398
if __FILE__ == $0
  generator = G5KHardwareGenerator.new("Hardware")
399

Lucas Nussbaum's avatar
Lucas Nussbaum committed
400 401 402 403 404 405 406 407 408 409 410 411 412 413 414
  options = WikiGenerator::parse_options
  if (options)
    ret = 2
    begin
      ret = generator.exec(options)
    rescue MediawikiApi::ApiError => e
      puts e, e.backtrace
      ret = 3
    rescue StandardError => e
      puts "Error with node: #{generator.instance_variable_get(:@node)}"
      puts e, e.backtrace
      ret = 4
    ensure
      exit(ret)
    end
415 416
  end
end