Identifying all cross joins between namespaces and users

Addressing Issue #415196

Changelog: other

db/gitlab_schemas/gitlab_main_cell.yaml

@@ -5,7 +5,7 @@ allow_cross_joins:
   - gitlab_main
   # Temporarily allow cross-joins between clusterwide and cell schemas
   # This is to be removed once we annotate all cross-joins between those
-  - gitlab_main_clusterwide
+  # - gitlab_main_clusterwide temporarily disabling this to discover all the cross-join violations
 allow_cross_transactions:
   - gitlab_internal
   - gitlab_shared

spec/support/database/prevent_cross_joins.rb

@@ -43,6 +43,25 @@ def self.validate_cross_joins!(sql)
 
       unless ::Gitlab::Database::GitlabSchema.cross_joins_allowed?(schemas)
         Thread.current[:has_cross_join_exception] = true
+        if (%w[users namespaces] - tables).empty?
+          open('rspec/cross_join_errors.txt', 'a') do |file| # this file will be in the tests artifacts
+            app_backtrace = caller.grep(/\/app\//)
+            spec_backtrace = caller.grep(/\_spec.rb/)
+            line = sql.match(/line:([^\*]+)/)[1]
+            # Hashing the item to deduplicate them
+            md5 = Digest::MD5.new
+            md5.update line # we can use another elements later for deduplication of items
+            hash = md5.hexdigest
+            item = {
+              hash: hash,
+              query: sql,
+              line: line,
+              app_backtrace: app_backtrace,
+              spec_backtrace: spec_backtrace
+            }
+            file.puts(item.to_json)
+          end
+        end
         raise CrossJoinAcrossUnsupportedTablesError,
           "Unsupported cross-join across '#{tables.join(", ")}' querying '#{schemas.to_a.join(", ")}' discovered " \
           "when executing query '#{sql}'. Please refer to https://docs.gitlab.com/ee/development/database/multiple_databases.html#removing-joins-between-ci_-and-non-ci_-tables for details on how to resolve this exception."