Hello Samuele,
> With your patch a refreshed view of the system and a task scheduling
> decision would be taken once every 5 seconds, no matter what. This
> could introduce unnecessary delays especially in the case of a system
> having many record updates.
So you agree that «maybe it was an overoptimization»? ;-)
I have reviewed the patches I'm applying in my 1.1 test instances and
the only one that I see that could be related to bibsched is this
trivial one:
diff --git a/lib/python/invenio/bibtask.py b/lib/python/invenio/bibtask.py
index f2a81b9..d2ad66f 100644
--- a/lib/python/invenio/bibtask.py
+++ b/lib/python/invenio/bibtask.py
@@ -1095,6 +1095,7 @@ def check_running_process_user():
configured in CFG_BIBSCHED_PROCESS_USER or as the user running the
Apache webserver process.
"""
+ return # Disable check at UAB
running_as_user = pwd.getpwuid(os.getuid())[0]
if CFG_BIBSCHED_PROCESS_USER:
# We have the expected bibsched process user defined in config,
As I'm running my Invenios as a normal Linux user, the
check_running_process_user() fails to do its work, giving false alarms
and aborting, so I just return. Otherwise, I cannot see any other
related modification.
The fact is that in my systems there is no sleep at all unless I move
the time.sleep() function as my patch indicates. Maybe you can see a
better place for this function, or maybe an i_have_slept_already boolean
variable or so?
Has anybody else appreciated this high CPU for bibsched usage in 1.1?
Thanks,
Ferran
