20735: Merge branch 'main' into 20735-update-deps

[arvados.git] / tools / crunchstat-summary / crunchstat_summary / summarizer.py

diff --git a/tools/crunchstat-summary/crunchstat_summary/summarizer.py b/tools/crunchstat-summary/crunchstat_summary/summarizer.py
index d99d3c1cf8324c4dd33f9bcdd72a15fa4f015d7c..a876257abc2a5ae00c8def17f4ae23e0219de446 100644 (file)
--- a/tools/crunchstat-summary/crunchstat_summary/summarizer.py
+++ b/tools/crunchstat-summary/crunchstat_summary/summarizer.py
@@ -105,7 +105,7 @@ class Summarizer(object):
                                        self.label, uuid)
                         continue
                     logger.debug('%s: follow %s', self.label, uuid)
-                    child_summarizer = ProcessSummarizer(uuid)
+                    child_summarizer = NewSummarizer(uuid)
                     child_summarizer.stats_max = self.stats_max
                     child_summarizer.task_stats = self.task_stats
                     child_summarizer.tasks = self.tasks
@@ -245,6 +245,27 @@ class Summarizer(object):
                     self.job_tot[category][stat] += val
         logger.debug('%s: done totals', self.label)
 
+        missing_category = {
+            'cpu': 'CPU',
+            'mem': 'memory',
+            'net:': 'network I/O',
+            'statfs': 'storage space',
+        }
+        for task_stat in self.task_stats.values():
+            for category in task_stat.keys():
+                for checkcat in missing_category:
+                    if checkcat.endswith(':'):
+                        if category.startswith(checkcat):
+                            missing_category.pop(checkcat)
+                            break
+                    else:
+                        if category == checkcat:
+                            missing_category.pop(checkcat)
+                            break
+        for catlabel in missing_category.values():
+            logger.warning('%s: %s stats are missing -- possible cluster configuration issue',
+                        self.label, catlabel)
+
     def long_label(self):
         label = self.label
         if hasattr(self, 'process') and self.process['uuid'] not in label:
@@ -364,7 +385,7 @@ class Summarizer(object):
                 constraint_key,
                 int(used_cores))
 
-    # FIXME: This needs to be updated to account for current nodemanager algorithms
+    # FIXME: This needs to be updated to account for current a-d-c algorithms
     def _recommend_ram(self):
         """Recommend an economical RAM constraint for this job.
 
@@ -526,7 +547,7 @@ def NewSummarizer(process_or_uuid, **kwargs):
 
 
 class ProcessSummarizer(Summarizer):
-    """Process is a job, pipeline, container, or container request."""
+    """Process is a job, pipeline, or container request."""
 
     def __init__(self, process, label=None, **kwargs):
         rdr = None
@@ -535,17 +556,16 @@ class ProcessSummarizer(Summarizer):
             label = self.process.get('name', self.process['uuid'])
         # Pre-Arvados v1.4 everything is in 'log'
         # For 1.4+ containers have no logs and container_requests have them in 'log_uuid', not 'log'
-        log_collection = self.process.get('log')
-        if not log_collection:
-            log_collection = self.process.get('log_uuid')
-        if log_collection:
+        log_collection = self.process.get('log', self.process.get('log_uuid'))
+        if log_collection and self.process.get('state') != 'Uncommitted': # arvados.util.CR_UNCOMMITTED:
             try:
                 rdr = crunchstat_summary.reader.CollectionReader(log_collection)
             except arvados.errors.NotFoundError as e:
                 logger.warning("Trying event logs after failing to read "
                                "log collection %s: %s", self.process['log'], e)
         if rdr is None:
-            rdr = crunchstat_summary.reader.LiveLogReader(self.process['uuid'])
+            uuid = self.process.get('container_uuid', self.process.get('uuid'))
+            rdr = crunchstat_summary.reader.LiveLogReader(uuid)
             label = label + ' (partial)'
         super(ProcessSummarizer, self).__init__(rdr, label=label, **kwargs)
         self.existing_constraints = self.process.get('runtime_constraints', {})