Merge branch 'sh-disable-cache-classes' into 'master'

Turn on caching of classes in Knapsack specs

See merge request !10709

.gitlab-ci.yml

@@ -67,6 +67,7 @@ stages:
     - export CI_NODE_TOTAL=${JOB_NAME[2]}
     - export KNAPSACK_REPORT_PATH=knapsack/${CI_PROJECT_NAME}/${JOB_NAME[0]}_node_${CI_NODE_INDEX}_${CI_NODE_TOTAL}_report.json
     - export KNAPSACK_GENERATE_REPORT=true
+    - export CACHE_CLASSES=true
     - cp ${KNAPSACK_RSPEC_SUITE_REPORT_PATH} ${KNAPSACK_REPORT_PATH}
     - knapsack rspec "--color --format documentation"
   artifacts:
@@ -87,6 +88,7 @@ stages:
     - export CI_NODE_TOTAL=${JOB_NAME[2]}
     - export KNAPSACK_REPORT_PATH=knapsack/${CI_PROJECT_NAME}/${JOB_NAME[0]}_node_${CI_NODE_INDEX}_${CI_NODE_TOTAL}_report.json
     - export KNAPSACK_GENERATE_REPORT=true
+    - export CACHE_CLASSES=true
     - cp ${KNAPSACK_SPINACH_SUITE_REPORT_PATH} ${KNAPSACK_REPORT_PATH}
     - knapsack spinach "-r rerun" || retry '[[ -e tmp/spinach-rerun.txt ]] && bundle exec spinach -r rerun $(cat tmp/spinach-rerun.txt)'
   artifacts:

config/environments/test.rb

@@ -8,7 +8,12 @@ Rails.application.configure do
   # test suite. You never need to work with it otherwise. Remember that
   # your test database is "scratch space" for the test suite and is wiped
   # and recreated between test runs. Don't rely on the data there!
-  config.cache_classes = false
+
+  # Enabling caching of classes slows start-up time because all controllers
+  # are loaded at initalization, but it reduces memory and load because files
+  # are not reloaded with every request. For example, caching is not necessary
+  # for loading database migrations but useful for handling Knapsack specs.
+  config.cache_classes = ENV['CACHE_CLASSES'] == 'true'
   # Configure static asset server for tests with Cache-Control for performance
   config.assets.digest = false

[Luke "Jared" Bennett]

Is this what was causing the memory issues recently?

[Stan Hu]

@lbennett No, but this should help reduce memory and increase the performance of our tests in general. I am now starting to suspect that the upgrade to gRPC actually may have been to blame for the increase memory usage. See https://gitlab.com/gitlab-org/gitlab-ee/issues/2162#note_27505657.

[Luke "Jared" Bennett]

Nice find @stanhu! 🤗