Implement job priorities in daily report (#100)

* Add first version of job priorities calculation (WIP) Signed-off-by: Crola1702 <[email protected]> * Add gazebo jobs priorities Signed-off-by: Crola1702 <[email protected]> * Return job priority in known issues section Signed-off-by: Crola1702 <[email protected]> * Show formatted priority in a new column Signed-off-by: Crola1702 <[email protected]> * Sort known issues by priority Signed-off-by: Crola1702 <[email protected]> * Rename sql query for clarity Signed-off-by: Crola1702 <[email protected]> * Address PR comments Signed-off-by: Crola1702 <[email protected]> --------- Signed-off-by: Crola1702 <[email protected]>
osrf · Nov 6, 2024 · 5bba611 · 5bba611
1 parent 377fadf
commit 5bba611
Show file tree

Hide file tree

Showing 4 changed files with 383 additions and 3 deletions.
diff --git a/database/scripts/get_known_issue_by_url.sql b/database/scripts/get_known_issue_by_url.sql
@@ -0,0 +1,6 @@
+SELECT job_name,
+    error_name
+FROM test_fail_issues
+WHERE github_issue = "@param1@"
+GROUP BY github_issue,
+    job_name;
diff --git a/database/scripts/lib/buildfarm_tools.rb b/database/scripts/lib/buildfarm_tools.rb
@@ -1,6 +1,7 @@
 # frozen_string_literal: true
 
 require 'open3'
+require 'csv'
 
 module BuildfarmToolsLib
   class BuildfarmToolsError < RuntimeError; end
@@ -15,6 +16,8 @@ class BuildfarmToolsError < RuntimeError; end
   FLAKY_BUILDS_DEFAULT_RANGE = '15 days'
   WARNING_AGE_CONSTANT = -1
 
+  JOB_PRIORITIES = CSV.read('lib/job_priorities.csv', converters: :numeric).to_h
+
   def self.build_regressions_today(filter_known: false)
     # Keys: job_name, build_number, build_datetime, failure_reason, last_section
     out = run_command('./sql_run.sh builds_failing_today.sql')
@@ -128,13 +131,49 @@ def self.jobs_last_success_date(older_than_days: 0)
     out
   end
 
-  def self.test_regressions_known
+  def self.test_regressions_known(sort_by: 'priority')
     out = known_issues(status: 'open')
     out.concat known_issues(status: 'disabled')
     out = out.group_by { |e| e["github_issue"] }.to_a.map { |e| e[1] }
+    out.each do |error_list|
+      priority = calculate_issue_priority(error_list.first["github_issue"])
+      error_list.each do |error|
+        error["priority"] = priority
+      end
+    end
+
+    unless sort_by.nil?
+      out.sort_by! { |r| -r.first['priority'] }
+    end
     out
   end
 
+  def self.calculate_issue_priority(issue_link)
+    sql_out = run_command('./sql_run.sh get_known_issue_by_url.sql', args: [issue_link])
+    errors = sql_out.map {|e| e['error_name']}.uniq
+    jobs = sql_out.map {|e| e['job_name']}.uniq
+
+    error_score_jobs = {}
+
+    errors.each do |e|
+      jobs.each do |job|
+        flaky_result = run_command('./sql_run.sh calculate_flakiness_jobs.sql', args: [e, FLAKY_BUILDS_DEFAULT_RANGE, job])
+        next if flaky_result.empty?
+        # This is not guaranteed to be 'not consistent', we need to re-check if the last 3 builds were failing because of this
+        flaky_ratio = flaky_result.first['failure_percentage'].to_f/100.0
+
+        job_priority = JOB_PRIORITIES[job]
+        job_priority = job_priority*1.5 if flaky_ratio == 1
+
+        error_score_jobs[job] = [] if error_score_jobs[job].nil?
+        error_score_jobs[job] << (job_priority*flaky_ratio)
+      end
+    end
+
+    # Get only maximum score for each job
+    error_score_jobs.each_value.map {|e| e.max}.sum.round(3)
+  end
+
   def self.run_command(cmd, args: [], keys: [])
     cmd += " '#{args.shift}'" until args.empty?
     begin