[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference noted::decps

Title:	POLYCENTER Performance Data Collector & Advisor for OpenVMS
Notice:	Latest version V2.2-51
Moderator:	BSS::JILSON

Created:	Wed Nov 07 1990
Last Modified:	Thu Jun 05 1997
Last Successful Update:	Fri Jun 06 1997
Number of topics:	2413
Total number of notes:	9017

2406.0. "How to specify phys cpu id in PSPA" by DV780::BAILEYR (Randy Baileyr@mail.dec.com) Tue Apr 22 1997 22:36

My customer site has a CI VAXCluster with 5 nodes, each of which
is a 7830.  When I try to generate a graph in csv format of CPU Busy
Interrupt I only receive the date of the sample and nothing more:

$ advise perf graph-
 /type=custom=(cpu_metrics=(p_busy,p_interrupt),select=0)-
 /node=node4-
 /format=csv-
 /begin=1-feb-1997-
 /end=2-feb-1997-
 /out=node4cpu0.csv

I've tried specifying physical cpu id's of "0", "1" and "2", but
all return no data, except the date/time of the sample.  I even
tried enclosing the cpu id in quotes.

I can get  CPU0 data using Datatrieve and the DTR record definition
for PSDC data, but I have to say:
DTR> find all node4cpu with cpu_id containing "0"  (vs. = "0")
                                   ^^^^^^^^^^^^^
Am I missing something?
DECPS-DC & DECPS-PA v2.1
OpenVMS 6.1
TIA
Randy

T.R	Title	User	Personal Name	Date	Lines
2406.1	needs node name prefix	OSOV03::KAGEYAMA	I Got Rhythm	`Wed Apr 23 1997 10:49`	7
	>$ advise perf graph- > /type=custom=(cpu_metrics=(p_busy,p_interrupt),select=0)- use /select=node4-0 - Kazunori
2406.2	nodename-0 worked - but...	DV780::BAILEYR	Randy Baileyr@mail.dec.com	`Wed Apr 23 1997 14:14`	7
	That did the trick! However, a new question. When I ran p_busy on node4-0 it returned values > 100%. Am I looking at CPU 0 only, or all 3 CPU's? When I ran the same graph on node4-1, it returned values < 100% for p_busy, which is what I would expect. Randy
2406.3		OSOV03::KAGEYAMA	I Got Rhythm	`Thu Apr 24 1997 02:37`	11
	> That did the trick! However, a new question. When I ran > p_busy on node4-0 it returned values > 100%. That's strange. It's the utilization of a given processor. So it should not exceed 100%(At least I believe). What's the value of p_idle at the same time. And also check the dump records to find the cause. $ ADVISE COLLECT REPORT DUMP_CPUS... - Kazunori
2406.4	use /NOSTACK	OSOV03::KAGEYAMA	I Got Rhythm	`Thu Apr 24 1997 03:10`	9
	> That did the trick! However, a new question. When I ran > p_busy on node4-0 it returned values > 100%. Have you combined it with other cpu metrics as you written in .0? It might be stacked or combined with other metrics. Then use /NOSTACK qualifier. /STACK is the default. - Kazunori
2406.5	/nostack is what I needed	DV780::BAILEYR	Randy Baileyr@mail.dec.com	`Thu Apr 24 1997 19:39`	13
	>Have you combined it with other cpu metrics as you written in .0? >It might be stacked or combined with other metrics. Then use /NOSTACK >qualifier. /STACK is the default. Yes, I was using (p_busy,p_interrupt). /nostack works great (now that I know)! tnks Randy