Don't recommend RAM or CPU changes based on zero usage. refs #10570

Never recommend 0 MB RAM. If CPU usage is zero, don't make any
CPU recommendations.

Arvados-DCO-1.1-Signed-off-by: Tom Morris <tfmorris@veritasgenetics.com>

diff --git a/tools/crunchstat-summary/crunchstat_summary/summarizer.py b/tools/crunchstat-summary/crunchstat_summary/summarizer.py
index 51f7e43e07a98c3f3908db0d6331bf41c22bbe5e..6d1fd918e4c9ab3df68643cb8d2bc25b7f92f5d6 100644 (file)
--- a/tools/crunchstat-summary/crunchstat_summary/summarizer.py
+++ b/tools/crunchstat-summary/crunchstat_summary/summarizer.py
@@ -323,6 +323,7 @@ class Summarizer(object):
             yield "# "+format_string.format(self._format(val))
 
     def _recommend_gen(self):
+        # TODO recommend fixing job granularity if elapsed time is too short
         return itertools.chain(
             self._recommend_cpu(),
             self._recommend_ram(),
@@ -333,9 +334,11 @@ class Summarizer(object):
 
         constraint_key = self._map_runtime_constraint('vcpus')
         cpu_max_rate = self.stats_max['cpu']['user+sys__rate']
-        if cpu_max_rate == float('-Inf'):
+        if cpu_max_rate == float('-Inf') or cpu_max_rate == 0.0:
             logger.warning('%s: no CPU usage data', self.label)
             return
+        # TODO Don't necessarily want to recommend on isolated max peak
+        # take average CPU usage into account as well or % time at max
         used_cores = max(1, int(math.ceil(cpu_max_rate)))
         asked_cores = self.existing_constraints.get(constraint_key)
         if asked_cores is None or used_cores < asked_cores:
@@ -391,8 +394,8 @@ class Summarizer(object):
         asked_mib = self.existing_constraints.get(constraint_key)
 
         nearlygibs = lambda mebibytes: mebibytes/AVAILABLE_RAM_RATIO/1024
-        if asked_mib is None or (
-                math.ceil(nearlygibs(used_mib)) < nearlygibs(asked_mib)):
+        if used_mib > 0 and (asked_mib is None or (
+                math.ceil(nearlygibs(used_mib)) < nearlygibs(asked_mib))):
             yield (
                 '#!! {} max RSS was {} MiB -- '
                 'try runtime_constraints "{}":{}'

diff --git a/tools/crunchstat-summary/tests/container_9tee4-dz642-lymtndkpy39eibk-arv-mount.txt.gz.report b/tools/crunchstat-summary/tests/container_9tee4-dz642-lymtndkpy39eibk-arv-mount.txt.gz.report
index 5148a4523af6c8a06cdc3568da5bc4272c1d1da1..0691e4f1ef4ea7e1604a0b7b73787f25b7dd7e58 100644 (file)
--- a/tools/crunchstat-summary/tests/container_9tee4-dz642-lymtndkpy39eibk-arv-mount.txt.gz.report
+++ b/tools/crunchstat-summary/tests/container_9tee4-dz642-lymtndkpy39eibk-arv-mount.txt.gz.report
@@ -20,5 +20,3 @@ time  elapsed 10      -       10
 # Max network speed in a single interval: 0.00MB/s
 # Keep cache miss rate 0.00%
 # Keep cache utilization 0.00%
-#!! container max CPU usage was 0% -- try runtime_constraints "vcpus":1
-#!! container max RSS was 0 MiB -- try runtime_constraints "ram":0

diff --git a/tools/crunchstat-summary/tests/logfile_20151210063411.txt.gz.report b/tools/crunchstat-summary/tests/logfile_20151210063411.txt.gz.report
index f9a34cfb98c7c6d42a633166b946b7c49db77cc1..c8b677450841b600586cfc5365a4c087708072ef 100644 (file)
--- a/tools/crunchstat-summary/tests/logfile_20151210063411.txt.gz.report
+++ b/tools/crunchstat-summary/tests/logfile_20151210063411.txt.gz.report
@@ -20,5 +20,4 @@ time  elapsed 2       -       4
 # Max network speed in a single interval: 0.00MB/s
 # Keep cache miss rate 0.00%
 # Keep cache utilization 0.00%
-#!! 4xphq-8i9sb-zvb2ocfycpomrup max CPU usage was 0% -- try runtime_constraints "min_cores_per_node":1
 #!! 4xphq-8i9sb-zvb2ocfycpomrup max RSS was 1 MiB -- try runtime_constraints "min_ram_mb_per_node":972

diff --git a/tools/crunchstat-summary/tests/logfile_20151210063439.txt.gz.report b/tools/crunchstat-summary/tests/logfile_20151210063439.txt.gz.report
index c54102d78a25d1158ba563a22de7279b3d39c2f2..b138b189313c9297a2f02d7bac07e6b14e2cd03a 100644 (file)
--- a/tools/crunchstat-summary/tests/logfile_20151210063439.txt.gz.report
+++ b/tools/crunchstat-summary/tests/logfile_20151210063439.txt.gz.report
@@ -20,5 +20,4 @@ time  elapsed 2       -       3
 # Max network speed in a single interval: 0.00MB/s
 # Keep cache miss rate 0.00%
 # Keep cache utilization 0.00%
-#!! 4xphq-8i9sb-v831jm2uq0g2g9x max CPU usage was 0% -- try runtime_constraints "min_cores_per_node":1
 #!! 4xphq-8i9sb-v831jm2uq0g2g9x max RSS was 1 MiB -- try runtime_constraints "min_ram_mb_per_node":972