X-Git-Url: https://git.arvados.org/arvados.git/blobdiff_plain/2540a26167950a6d80f5eaf183500f22739d17de..3dc2b0008be4e09b47f73aeeb97d005cd7836619:/services/api/script/crunch-dispatch.rb diff --git a/services/api/script/crunch-dispatch.rb b/services/api/script/crunch-dispatch.rb index a9ca4bebc4..a23b02b15c 100755 --- a/services/api/script/crunch-dispatch.rb +++ b/services/api/script/crunch-dispatch.rb @@ -2,6 +2,7 @@ include Process +$warned = {} $signal = {} %w{TERM INT}.each do |sig| signame = sig @@ -26,6 +27,7 @@ require File.dirname(__FILE__) + '/../config/environment' require 'open3' $redis ||= Redis.new +LOG_BUFFER_SIZE = 2**20 class Dispatcher include ApplicationHelper @@ -126,6 +128,10 @@ class Dispatcher cmd_args << '--job' cmd_args << job.uuid + if cmd_args[0] == '' + raise "No CRUNCH_JOB_BIN env var, and crunch-job not in path." + end + commit = Commit.where(sha1: job.script_version).first if commit cmd_args << '--git-dir' @@ -152,8 +158,14 @@ class Dispatcher untake(job) next end - $stderr.puts "dispatch: job #{job.uuid} start" - $stderr.puts "dispatch: child #{t.pid} start" + + $stderr.puts "dispatch: job #{job.uuid}" + start_banner = "dispatch: child #{t.pid} start #{Time.now.ctime.to_s}" + $stderr.puts start_banner + $redis.set job.uuid, start_banner + "\n" + $redis.publish job.uuid, start_banner + $redis.publish job.owner_uuid, start_banner + @running[job.uuid] = { stdin: i, stdout: o, @@ -208,7 +220,16 @@ class Dispatcher lines.each do |line| $stderr.print "#{job_uuid} ! " unless line.index(job_uuid) $stderr.puts line - $redis.publish job_uuid, "#{Time.now.ctime.to_s} #{line.strip}" + pub_msg = "#{Time.now.ctime.to_s} #{line.strip}" + $redis.publish job.owner_uuid, pub_msg + $redis.publish job_uuid, pub_msg + $redis.append job_uuid, pub_msg + "\n" + if LOG_BUFFER_SIZE < $redis.strlen(job_uuid) + $redis.set(job_uuid, + $redis + .getrange(job_uuid, (LOG_BUFFER_SIZE >> 1), -1) + .sub(/^.*?\n/, '')) + end end end end @@ -291,7 +312,9 @@ class Dispatcher else refresh_todo unless did_recently(:refresh_todo, 1.0) update_node_status - start_jobs unless @todo.empty? or did_recently(:start_jobs, 1.0) + unless @todo.empty? or did_recently(:start_jobs, 1.0) or $signal[:term] + start_jobs + end end reap_children select(@running.values.collect { |j| [j[:stdout], j[:stderr]] }.flatten, @@ -312,4 +335,7 @@ class Dispatcher end end +# This is how crunch-job child procs know where the "refresh" trigger file is +ENV["CRUNCH_REFRESH_TRIGGER"] = Rails.configuration.crunch_refresh_trigger + Dispatcher.new.run