switchman-inst-jobs 3.0.3 → 3.1.2

checksums.yaml

@@ -1,7 +1,7 @@
 ---
 SHA256:
-  metadata.gz: '08102eee5c1da6d032170565b624d61f90fcbc1c6232e1214babda846c9674c9'
-  data.tar.gz: 5f9c75d442f40b706dff2a45bfab8dc99ff8599470c096bbea594f1a763d7640
+  metadata.gz: 4efbe05e879c9162df06f945af688c64876b9c56cc870509687160549359fd39
+  data.tar.gz: e32c8a6dfecd0b6c760d07a4d45eeaf6651d6cc7e70e7b6418d82c5a112f70cd
 SHA512:
-  metadata.gz: 9118a73e2a47f1e6083737f967cc298674b8884c8cc23c9d16efff9e087951d63e54129f51a1c5595f2ec4cadb4332208a2a0262e7992b5777c39ac08ae8efa1
-  data.tar.gz: 791c382a9fbb141dc0e2f763b936d953bbd575ac01fc869a652628992c93b47c3cf93b590d95bd7867698c9410cbb278932ed1404e1e6d9f582cf00825cda8b0
+  metadata.gz: 4fe0103d1b4bc305f0c85ba2135366968eb7c33bb6482b6d302dde9b9532db1d3d09fac64d90dad16fe6c6a438236ede5732d16a22d8538053d814010c9d2c87
+  data.tar.gz: feb2e546ee06bc38f964f58e63cfb59fdc25757952762ef8f8faf0521050d264e9f018ea8e6bb0131b277997d63035f68562cf45a7af13886fbfd80f481f5515

data/db/migrate/20110610213249_optimize_delayed_jobs.rb

@@ -29,7 +29,7 @@ class OptimizeDelayedJobs < ActiveRecord::Migration[4.2]
     add_index :delayed_jobs, %w[strand id], name: 'index_delayed_jobs_on_strand'
 
     # move all failed jobs to the new failed table
-    Delayed::Backend::ActiveRecord::Job.where('failed_at IS NOT NULL').find_each do |job|
+    Delayed::Backend::ActiveRecord::Job.where.not(failed_at: nil).find_each do |job|
       job.fail! unless job.on_hold?
     end
   end

data/lib/switchman_inst_jobs.rb

@@ -38,6 +38,7 @@ end
 
 require 'switchman_inst_jobs/active_record/connection_adapters/postgresql_adapter'
 require 'switchman_inst_jobs/active_record/migration'
+require 'switchman_inst_jobs/delayed/settings'
 require 'switchman_inst_jobs/delayed/backend/base'
 require 'switchman_inst_jobs/delayed/message_sending'
 require 'switchman_inst_jobs/delayed/pool'

data/lib/switchman_inst_jobs/delayed/backend/base.rb

@@ -43,8 +43,7 @@ module SwitchmanInstJobs
           end
 
           def configured_shard_ids
-            (::Delayed::Settings.worker_config.try(:[], 'workers') || []).
-              map { |w| w['shard'] }.compact.uniq
+            ::SwitchmanInstJobs::Delayed::Settings.configured_shard_ids
           end
 
           def processes_locked_locally
@@ -93,7 +92,7 @@ module SwitchmanInstJobs
           # likely a missing shard with a stale cache
           current_shard.send(:clear_cache)
           ::Switchman::Shard.clear_cache
-          raise ShardNotFoundError, shard_id unless ::Switchman::Shard.where(id: shard_id).exists?
+          raise ShardNotFoundError, shard_id unless ::Switchman::Shard.exists?(id: shard_id)
 
           raise
         end

data/lib/switchman_inst_jobs/delayed/pool.rb

@@ -17,7 +17,7 @@ module SwitchmanInstJobs
           # We purposely don't .compact to remove nils here, since if any
           # workers are on the default jobs shard we want to unlock against
           # that shard too.
-          shard_ids = @config[:workers].map { |c| c[:shard] }.uniq
+          shard_ids = @config[:workers].pluck(:shard).uniq
           shards = shard_ids.map { |shard_id| ::Delayed::Worker.shard(shard_id) }
         end
         ::Switchman::Shard.with_each_shard(shards, [:delayed_jobs]) do

data/lib/switchman_inst_jobs/delayed/settings.rb

@@ -0,0 +1,9 @@
+module SwitchmanInstJobs
+  module Delayed
+    module Settings
+      def self.configured_shard_ids
+        (::Delayed::Settings.worker_config.try(:[], 'workers') || []).map { |w| w['shard'] }.compact.uniq
+      end
+    end
+  end
+end

data/lib/switchman_inst_jobs/delayed/worker/health_check.rb

@@ -18,22 +18,23 @@ module SwitchmanInstJobs
             ::Delayed::Settings.worker_health_check_config['service_name'] = original_service_name
           end
 
-          def reschedule_abandoned_jobs(call_super: false)
-            shards = ::Switchman::Shard.delayed_jobs_shards.to_a
-            call_super = shards.first if shards.length == 1
-            unless call_super == false
-              call_super.activate(:delayed_jobs) do
-                return munge_service_name(call_super) { super() }
+          def reschedule_abandoned_jobs
+            shard_ids = ::SwitchmanInstJobs::Delayed::Settings.configured_shard_ids
+            shards = shard_ids.map { |shard_id| ::Delayed::Worker.shard(shard_id) }
+            ::Switchman::Shard.with_each_shard(shards, [:delayed_jobs]) do
+              dj_shard = ::Switchman::Shard.current(:delayed_jobs)
+              dj_shard.activate do
+                munge_service_name(dj_shard) do
+                  # because this rescheduling process is running on every host, we need
+                  # to make sure that it's functioning for each shard the current
+                  # host is programmed to interact with, but ONLY for those shards.
+                  # reading the config lets us iterate over any shards this host should
+                  # work with and lets us pick the correct service name to identify which
+                  # hosts are currently alive and valid via the health checks
+                  super()
+                end
               end
             end
-
-            ::Switchman::Shard.with_each_shard(shards, [:delayed_jobs], exception: :ignore) do
-              shard = ::Switchman::Shard.current(:delayed_jobs)
-              singleton = <<~SINGLETON
-                periodic: Delayed::Worker::HealthCheck.reschedule_abandoned_jobs:#{shard.id}
-              SINGLETON
-              delay(singleton: singleton).reschedule_abandoned_jobs(call_super: shard)
-            end
           end
         end
       end

data/lib/switchman_inst_jobs/jobs_migrator.rb

@@ -29,26 +29,59 @@ module SwitchmanInstJobs
 
       def migrate_shards(shard_map)
         source_shards = Set[]
+        target_shards = Hash.new([])
         shard_map.each do |(shard, target_shard)|
           shard = ::Switchman::Shard.find(shard) unless shard.is_a?(::Switchman::Shard)
           source_shards << shard.delayed_jobs_shard.id
-          # If target_shard is an int, it won't have an id, but we can just use it as is
-          shard.update(delayed_jobs_shard_id: target_shard.try(:id) || target_shard, block_stranded: true)
+          target_shard = target_shard.try(:id) || target_shard
+          target_shards[target_shard] += [shard.id]
         end
 
+        # Do the updates in batches and then just clear redis instead of clearing them one at a time
+        target_shards.each do |target_shard, shards|
+          updates = { delayed_jobs_shard_id: target_shard, block_stranded: true }
+          updates[:updated_at] = Time.zone.now if ::Switchman::Shard.column_names.include?('updated_at')
+          ::Switchman::Shard.where(id: shards).update_all(updates)
+        end
+        clear_shard_cache
+
+        ::Switchman::Shard.clear_cache
+        # rubocop:disable Style/CombinableLoops
+        # We first migrate strands so that we can stop blocking strands before we migrate unstranded jobs
+        source_shards.each do |s|
+          ::Switchman::Shard.lookup(s).activate(:delayed_jobs) { migrate_strands }
+        end
+
+        source_shards.each do |s|
+          ::Switchman::Shard.lookup(s).activate(:delayed_jobs) { migrate_everything }
+        end
+        ensure_unblock_stranded_for(shard_map.map(&:first))
+        # rubocop:enable Style/CombinableLoops
+      end
+
+      # if :migrate_strands ran on any shards that fell into scenario 1, then
+      # block_stranded never got flipped, so do that now.
+      def ensure_unblock_stranded_for(shards)
+        shards = ::Switchman::Shard.where(id: shards, block_stranded: true).to_a
+        return unless shards.any?
+
+        ::Switchman::Shard.where(id: shards).update_all(block_stranded: false)
+        clear_shard_cache
+
+        # shards is an array of shard objects that is now stale cause block_stranded has been updated.
+        shards.map(&:delayed_jobs_shard).uniq.each do |dj_shard|
+          unblock_strands(dj_shard)
+        end
+      end
+
+      def clear_shard_cache(debug_message = nil)
+        ::Switchman.cache.clear
+        Rails.logger.debug("Waiting for caches to clear #{debug_message}")
         # Wait a little over the 60 second in-process shard cache clearing
         # threshold to ensure that all new stranded jobs are now being
         # enqueued with next_in_strand: false
-        Rails.logger.debug("Waiting for caches to clear (#{source_shard.id} -> #{target_shard.id})")
+        # @skip_cache_wait is for spec usage only
         sleep(65) unless @skip_cache_wait
-
-        # TODO: 4 has been picked completely out of a hat.  We should make it configurable or something
-        Parallel.each(source_shards, in_processes: 4) do |s|
-          # Ensure the child processes don't share connections with the parent
-          Delayed::Pool.on_fork.call
-          ActiveRecord::Base.clear_all_connections!
-          s.activate(:delayed_jobs) { run }
-        end
       end
 
       # This method expects that all relevant shards already have block_stranded: true
@@ -71,7 +104,7 @@ module SwitchmanInstJobs
         #    those (= do nothing since it should already be true)
 
         source_shard = ::Switchman::Shard.current(:delayed_jobs)
-        strand_scope = ::Delayed::Job.shard(source_shard).where('strand IS NOT NULL')
+        strand_scope = ::Delayed::Job.shard(source_shard).where.not(strand: nil)
         shard_map = build_shard_map(strand_scope, source_shard)
         shard_map.each do |(target_shard, source_shard_ids)|
           shard_scope = strand_scope.where(shard_id: source_shard_ids)
@@ -92,7 +125,7 @@ module SwitchmanInstJobs
                 # We lock it to ensure that the jobs worker can't delete it until we are done moving the strand
                 # Since we only unlock it on the new jobs queue *after* deleting from the original
                 # the lock ensures the blocker always gets unlocked
-                first = this_strand_scope.where('locked_by IS NOT NULL').next_in_strand_order.lock.first
+                first = this_strand_scope.where.not(locked_by: nil).next_in_strand_order.lock.first
                 if first
                   first_job = ::Delayed::Job.create!(strand: strand, next_in_strand: false)
                   first_job.payload_object = ::Delayed::PerformableMethod.new(Kernel, :sleep, args: [0])
@@ -110,7 +143,7 @@ module SwitchmanInstJobs
                     update_all(next_in_strand: false)
                 end
 
-                # 4) is taken care of here, by leaveing next_in_strand alone and
+                # 4) is taken care of here, by leaving next_in_strand alone and
                 # it should execute on the new shard
                 batch_move_jobs(
                   target_shard: target_shard,
@@ -123,27 +156,35 @@ module SwitchmanInstJobs
               end
             end
 
-            ::Switchman::Shard.find(source_shard_ids).each do |shard|
-              shard.update(block_stranded: false)
-            end
-            # Wait a little over the 60 second in-process shard cache clearing
-            # threshold to ensure that all new stranded jobs are now being
-            # enqueued with next_in_strand: false
-            Rails.logger.debug("Waiting for caches to clear (#{source_shard.id} -> #{target_shard.id})")
-            # for spec usage only
-            sleep(65) unless @skip_cache_wait
+            updated = ::Switchman::Shard.where(id: source_shard_ids, block_stranded: true).
+              update_all(block_stranded: false)
+            # If this is being manually re-run for some reason to clean something up, don't wait for nothing to happen
+            clear_shard_cache("(#{source_shard.id} -> #{target_shard.id})") unless updated.zero?
+
+            ::Switchman::Shard.clear_cache
             # At this time, let's unblock all the strands on the target shard that aren't being held by a blocker
             # but actually could have run and we just didn't know it because we didn't know if they had jobs
             # on the source shard
-            # rubocop:disable Layout/LineLength
-            strands_to_unblock = shard_scope.where.not(source: 'JobsMigrator::StrandBlocker').
-              distinct.
-              where("NOT EXISTS (SELECT 1 FROM #{::Delayed::Job.quoted_table_name} dj2 WHERE delayed_jobs.strand=dj2.strand AND next_in_strand)").
-              pluck(:strand)
-            # rubocop:enable Layout/LineLength
-            strands_to_unblock.each do |strand|
-              Delayed::Job.where(strand: strand).next_in_strand_order.first.update_attribute(:next_in_strand, true)
-            end
+            unblock_strands(target_shard)
+          end
+        end
+      end
+
+      def unblock_strands(target_shard)
+        target_shard.activate(:delayed_jobs) do
+          loop do
+            # We only want to unlock stranded jobs where they don't belong to a blocked shard (if they *do* belong)
+            # to a blocked shard, they must be part of a concurrent jobs migration from a different source shard to
+            # this target shard, so we shouldn't unlock them yet.  We only ever unlock one job here to keep the
+            # logic cleaner; if the job is n-stranded, after the first one runs, the trigger will unlock larger
+            # batches
+            break if ::Delayed::Job.where(id: ::Delayed::Job.select('DISTINCT ON (strand) id').
+              where.not(strand: nil).
+              where.not(shard_id: ::Switchman::Shard.where(block_stranded: true).pluck(:id)).where(
+                ::Delayed::Job.select(1).from("#{::Delayed::Job.quoted_table_name} dj2").
+                where("dj2.next_in_strand = true OR dj2.source = 'JobsMigrator::StrandBlocker'").
+                where('dj2.strand = delayed_jobs.strand').arel.exists.not
+              ).order(:strand, :strand_order_override, :id)).limit(500).update_all(next_in_strand: true).zero?
           end
         end
       end
@@ -242,7 +283,7 @@ module SwitchmanInstJobs
 
         connection.execute "COPY #{::Delayed::Job.quoted_table_name} (#{quoted_keys}) FROM STDIN"
         records.map do |record|
-          connection.raw_connection.put_copy_data(keys.map { |k| quote_text(record[k]) }.join("\t") + "\n")
+          connection.raw_connection.put_copy_data("#{keys.map { |k| quote_text(record[k]) }.join("\t")}\n")
         end
         connection.clear_query_cache
         connection.raw_connection.put_copy_end

data/lib/switchman_inst_jobs/switchman/shard.rb

@@ -14,7 +14,7 @@ module SwitchmanInstJobs
           shard = ::Switchman::Shard.lookup(delayed_jobs_shard_id)
           return shard if shard
         end
-        database_server&.delayed_jobs_shard(self)
+        @delayed_jobs_shard ||= database_server&.delayed_jobs_shard(self)
       end
 
       # Adapted from hold/unhold methods in base delayed jobs base
@@ -39,7 +39,14 @@ module SwitchmanInstJobs
 
       def unhold_jobs!
         self.jobs_held = false
-        save! if changed?
+        if changed?
+          save!
+          # Wait a little over the 60 second in-process shard cache clearing
+          # threshold to ensure that all new jobs are now being enqueued
+          # unlocked
+          Rails.logger.debug('Waiting for caches to clear')
+          sleep(65)
+        end
         delayed_jobs_shard.activate(:delayed_jobs) do
           ::Delayed::Job.where(locked_by: ::Delayed::Backend::Base::ON_HOLD_LOCKED_BY, shard_id: id).
             in_batches(of: 10_000).

data/lib/switchman_inst_jobs/timed_cache.rb

@@ -6,7 +6,7 @@ module SwitchmanInstJobs
       @cached_at = Time.zone.now
     end
 
-    def clear(force = false)
+    def clear(force: false)
       if force || @cached_at < @timeout.call
         @block.call
         @cached_at = Time.zone.now

data/lib/switchman_inst_jobs/version.rb

@@ -1,3 +1,3 @@
 module SwitchmanInstJobs
-  VERSION = '3.0.3'.freeze
+  VERSION = '3.1.2'.freeze
 end

metadata

@@ -1,35 +1,35 @@
 --- !ruby/object:Gem::Specification
 name: switchman-inst-jobs
 version: !ruby/object:Gem::Version
-  version: 3.0.3
+  version: 3.1.2
 platform: ruby
 authors:
 - Bryan Petty
-autorequire:
+autorequire: 
 bindir: bin
 cert_chain: []
-date: 2020-11-11 00:00:00.000000000 Z
+date: 2021-01-26 00:00:00.000000000 Z
 dependencies:
 - !ruby/object:Gem::Dependency
   name: inst-jobs
   requirement: !ruby/object:Gem::Requirement
     requirements:
-    - - "~>"
-      - !ruby/object:Gem::Version
-        version: '1.0'
     - - ">="
       - !ruby/object:Gem::Version
         version: 1.0.3
+    - - "<"
+      - !ruby/object:Gem::Version
+        version: '3.0'
   type: :runtime
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
-    - - "~>"
-      - !ruby/object:Gem::Version
-        version: '1.0'
     - - ">="
       - !ruby/object:Gem::Version
         version: 1.0.3
+    - - "<"
+      - !ruby/object:Gem::Version
+        version: '3.0'
 - !ruby/object:Gem::Dependency
   name: parallel
   requirement: !ruby/object:Gem::Requirement
@@ -210,28 +210,28 @@ dependencies:
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.79.0
+        version: 1.3.1
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 0.79.0
+        version: 1.3.1
 - !ruby/object:Gem::Dependency
   name: rubocop-rails
   requirement: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 2.4.2
+        version: 2.8.1
   type: :development
   prerelease: false
   version_requirements: !ruby/object:Gem::Requirement
     requirements:
     - - "~>"
       - !ruby/object:Gem::Version
-        version: 2.4.2
+        version: 2.8.1
 - !ruby/object:Gem::Dependency
   name: simplecov
   requirement: !ruby/object:Gem::Requirement
@@ -260,7 +260,7 @@ dependencies:
     - - "~>"
       - !ruby/object:Gem::Version
         version: '1.4'
-description:
+description: 
 email:
 - bpetty@instructure.com
 executables: []
@@ -303,6 +303,7 @@ files:
 - lib/switchman_inst_jobs/delayed/backend/base.rb
 - lib/switchman_inst_jobs/delayed/message_sending.rb
 - lib/switchman_inst_jobs/delayed/pool.rb
+- lib/switchman_inst_jobs/delayed/settings.rb
 - lib/switchman_inst_jobs/delayed/worker.rb
 - lib/switchman_inst_jobs/delayed/worker/health_check.rb
 - lib/switchman_inst_jobs/engine.rb
@@ -319,7 +320,7 @@ homepage: https://github.com/instructure/switchman-inst-jobs
 licenses:
 - MIT
 metadata: {}
-post_install_message:
+post_install_message: 
 rdoc_options: []
 require_paths:
 - lib
@@ -334,8 +335,8 @@ required_rubygems_version: !ruby/object:Gem::Requirement
     - !ruby/object:Gem::Version
       version: '0'
 requirements: []
-rubygems_version: 3.1.4
-signing_key:
+rubygems_version: 3.0.3
+signing_key: 
 specification_version: 4
 summary: Switchman and Instructure Jobs compatibility gem.
 test_files: []