https://bugzilla.redhat.com/show_bug.cgi?id=1377464
Bug ID: 1377464
Summary: "pmrep proc" takes too long
Product: Fedora
Version: rawhide
Component: pcp
Assignee: nathans@xxxxxxxxxx
Reporter: fche@xxxxxxxxxx
QA Contact: extras-qa@xxxxxxxxxxxxxxxxx
CC: brolley@xxxxxxxxxx, fche@xxxxxxxxxx, lberk@xxxxxxxxxx,
mgoodwin@xxxxxxxxxx, nathans@xxxxxxxxxx,
pcp@xxxxxxxxxxx, scox@xxxxxxxxxx
Marko reports "pmrep proc" takes on the order of many seconds. Diagnosis with
the aid of a systemtap script (ltrace and $PCP_DEBUG not being helpful)
indicated that the pmrep is using the extend_item rather than extend_indom API
for the metrics beneath proc.*, one instance per call.
This means that each extend_item involves another bunch of lookups, repeated
over processes (the indom), and over metrics (proc.*).
pmrep should routinely use the extend_indom API for metrics with indoms. (It
can use the same API for metrics with a null indom!). It can find the instance
domain in the output results coming from each fetch.
--
You are receiving this mail because:
You are on the CC list for the bug.
Unsubscribe from this bug
https://bugzilla.redhat.com/token.cgi?t=oq2zT9macB&a=cc_unsubscribe
|