13493: Update CORS OPTIONS test.

[arvados.git] / services / api / lib / crunch_dispatch.rb

diff --git a/services/api/lib/crunch_dispatch.rb b/services/api/lib/crunch_dispatch.rb
index 230f03e5777a56aec28b57bda56dcea4e82b065f..73ad7606cc879ef58f7569c960196191c7fb7721 100644 (file)
--- a/services/api/lib/crunch_dispatch.rb
+++ b/services/api/lib/crunch_dispatch.rb
@@ -297,7 +297,7 @@ class CrunchDispatch
     @fetched_commits[sha1] = ($? == 0)
   end
 
-  def tag_commit(commit_hash, tag_name)
+  def tag_commit(job, commit_hash, tag_name)
     # @git_tags[T]==V if we know commit V has been tagged T in the
     # arvados_internal repository.
     if not @git_tags[tag_name]
@@ -381,20 +381,20 @@ class CrunchDispatch
           next
         end
         ready &&= get_commit repo.server_path, job.script_version
-        ready &&= tag_commit job.script_version, job.uuid
+        ready &&= tag_commit job, job.script_version, job.uuid
       end
 
       # This should be unnecessary, because API server does it during
       # job create/update, but it's still not a bad idea to verify the
       # tag is correct before starting the job:
-      ready &&= tag_commit job.script_version, job.uuid
+      ready &&= tag_commit job, job.script_version, job.uuid
 
       # The arvados_sdk_version doesn't support use of arbitrary
       # remote URLs, so the requested version isn't necessarily copied
       # into the internal repository yet.
       if job.arvados_sdk_version
         ready &&= get_commit @arvados_repo_path, job.arvados_sdk_version
-        ready &&= tag_commit job.arvados_sdk_version, "#{job.uuid}-arvados-sdk"
+        ready &&= tag_commit job, job.arvados_sdk_version, "#{job.uuid}-arvados-sdk"
       end
 
       if not ready
@@ -429,8 +429,11 @@ class CrunchDispatch
         i, o, e, t = Open3.popen3(*cmd_args)
       rescue
         $stderr.puts "dispatch: popen3: #{$!}"
-        sleep 1
-        next
+        # This is a dispatch problem like "Too many open files";
+        # retrying another job right away would be futile. Just return
+        # and hope things are better next time, after (at least) a
+        # did_recently() delay.
+        return
       end
 
       $stderr.puts "dispatch: job #{job.uuid}"
@@ -633,31 +636,11 @@ class CrunchDispatch
     pid_done = nil
     j_done = nil
 
-    if false
-      begin
-        pid_done = waitpid(-1, Process::WNOHANG | Process::WUNTRACED)
-        if pid_done
-          j_done = @running.values.
-            select { |j| j[:wait_thr].pid == pid_done }.
-            first
-        end
-      rescue SystemCallError
-        # I have @running processes but system reports I have no
-        # children. This is likely to happen repeatedly if it happens at
-        # all; I will log this no more than once per child process I
-        # start.
-        if 0 < @running.select { |uuid,j| j[:warned_waitpid_error].nil? }.size
-          children = @running.values.collect { |j| j[:wait_thr].pid }.join ' '
-          $stderr.puts "dispatch: IPC bug: waitpid() error (#{$!}), but I have children #{children}"
-        end
-        @running.each do |uuid,j| j[:warned_waitpid_error] = true end
-      end
-    else
-      @running.each do |uuid, j|
-        if j[:wait_thr].status == false
-          pid_done = j[:wait_thr].pid
-          j_done = j
-        end
+    @running.each do |uuid, j|
+      if !j[:wait_thr].status
+        pid_done = j[:wait_thr].pid
+        j_done = j
+        break
       end
     end