Merge "add per job percentages for responsibility of fails"

elastic_recheck/cmd/check_success.py

@@ -15,6 +15,7 @@
 #    under the License.
 
 import argparse
+import collections
 import operator
 import os
 import re
@@ -63,7 +64,15 @@ def all_fails(classifier):
     return all_fails
 
 
-def classifying_rate(classifier, data):
+def num_fails_per_build_name(all_jobs):
+    counts = collections.defaultdict(int)
+    for f in all_jobs:
+        build, job = f.split('.', 1)
+        counts[job] += 1
+    return counts
+
+
+def classifying_rate(fails, data):
     """Builds and prints the classification rate.
 
     It's important to know how good a job we are doing, so this
@@ -71,24 +80,20 @@ def classifying_rate(classifier, data):
     classification rate. For every failure in the gate queue did
     we find a match for it.
     """
-    fails = all_fails(classifier)
     for bugnum in data:
         bug = data[bugnum]
         for job in bug['failed_jobs']:
             fails[job] = True
 
     total = len(fails.keys())
-    bad_jobs = {}
+    bad_jobs = collections.defaultdict(int)
     count = 0
     for f in fails:
         if fails[f] is True:
             count += 1
         else:
             build, job = f.split('.', 1)
-            if job in bad_jobs:
-                bad_jobs[job] += 1
-            else:
-                bad_jobs[job] = 1
+            bad_jobs[job] += 1
 
     print("Classification percentage: %2.2f%%" %
           ((float(count) / float(total)) * 100.0))
@@ -101,28 +106,73 @@ def classifying_rate(classifier, data):
         print "  %3s : %s" % (s[1], s[0])
 
 
-def collect_metrics(classifier):
+def _status_count(results):
+    counts = {}
+    facets = er_results.FacetSet()
+    facets.detect_facets(
+        results,
+        ["build_status", "build_uuid"])
+
+    for key in facets:
+        counts[key] = len(facets[key])
+    return counts
+
+
+def _failure_count(hits):
+    if "FAILURE" in hits:
+        return hits["FAILURE"]
+    else:
+        return 0
+
+
+def _failed_jobs(results):
+    failed_jobs = []
+    facets = er_results.FacetSet()
+    facets.detect_facets(
+        results,
+        ["build_status", "build_uuid"])
+    if "FAILURE" in facets:
+        for build in facets["FAILURE"]:
+            for result in facets["FAILURE"][build]:
+                failed_jobs.append("%s.%s" % (build, result.build_name))
+    return failed_jobs
+
+
+def _count_fails_per_build_name(hits):
+    facets = er_results.FacetSet()
+    counts = collections.defaultdict(int)
+    facets.detect_facets(
+        hits,
+        ["build_status", "build_name", "build_uuid"])
+    if "FAILURE" in facets:
+        for build_name in facets["FAILURE"]:
+            counts[build_name] += 1
+    return counts
+
+
+def _failure_percentage(hits, fails):
+    total_fails_per_build_name = num_fails_per_build_name(fails)
+    fails_per_build_name = _count_fails_per_build_name(hits)
+    per = {}
+    for build in fails_per_build_name:
+        this_job = fails_per_build_name[build]
+        if build in total_fails_per_build_name:
+            total = total_fails_per_build_name[build]
+            per[build] = (float(this_job) / float(total)) * 100.0
+    return per
+
+
+def collect_metrics(classifier, fails):
     data = {}
     for q in classifier.queries:
         results = classifier.hits_by_query(q['query'], size=30000)
-        facets = er_results.FacetSet()
-        facets.detect_facets(
-            results,
-            ["build_status", "build_uuid"])
-
-        num_fails = 0
-        failed_jobs = []
-        if "FAILURE" in facets:
-            num_fails = len(facets["FAILURE"])
-            for build in facets["FAILURE"]:
-                for result in facets["FAILURE"][build]:
-                    failed_jobs.append("%s.%s" % (build, result.build_name))
-
+        hits = _status_count(results)
         data[q['bug']] = {
-            'fails': num_fails,
-            'hits': facets,
+            'fails': _failure_count(hits),
+            'hits': hits,
+            'percentages': _failure_percentage(results, fails),
             'query': q['query'],
-            'failed_jobs': failed_jobs
+            'failed_jobs': _failed_jobs(results)
         }
 
     return data
@@ -135,13 +185,18 @@ def print_metrics(data, with_lp=False):
     sorted_data = sorted(data.iteritems(),
                          key=lambda x: -x[1]['fails'])
     for d in sorted_data:
+        bug = d[0]
+        data = d[1]
         print("Bug: https://bugs.launchpad.net/bugs/%s => %s"
-              % (d[0], d[1]['query'].rstrip()))
+              % (bug, data['query'].rstrip()))
         if with_lp:
             get_launchpad_bug(d[0])
         print "Hits"
-        for s in d[1]['hits'].keys():
-            print "  %s: %s" % (s, len(d[1]['hits'][s]))
+        for s in data['hits']:
+            print "  %s: %s" % (s, data['hits'][s])
+        print "Percentage of Gate Queue Job failures triggered by this bug"
+        for s in data['percentages']:
+            print "  %s: %2.2f%%" % (s, data['percentages'][s])
         print
 
 
@@ -160,10 +215,11 @@ def get_launchpad_bug(bug):
 def main():
     opts = get_options()
     classifier = er.Classifier(opts.dir)
-    data = collect_metrics(classifier)
+    fails = all_fails(classifier)
+    data = collect_metrics(classifier, fails)
     print_metrics(data, with_lp=opts.lp)
     if opts.rate:
-        classifying_rate(classifier, data)
+        classifying_rate(fails, data)
 
 
 if __name__ == "__main__":