|
I think you need to change the snmpCol.lrf to have
more concurrent collections.
Run
snmpCollect -S
and then view
/usr/OV/log/snmpCol.trace.
You'll probably see a line near the end of the file
like
1 of 5 concurrent SNMP requests in use
I'd suggest you can happily increase that
concurrent setting to say 50 would be OK.
Edit /usr/OV/lrf/snmpCol.lrf and change
it from :
OVs_YES_START:nvsecd,trapd,ovwdb,ovtopmd::OVs_WELL_BEHAVED:120:
to
OVs_YES_START:nvsecd,trapd,ovwdb,ovtopmd:-n
50:OVs_WELL_BEHAVED:120:
You then need to
ovaddobj snmpCol.lrf
ovstop snmpCollect
ovstart snmpCollect
snmpCollect -S
and again view /usr/OV/log/snmpCol.trace,
where you should see a later line like:
1 of 50
concurrent SNMP requests in use
It adds more network and CPU load to your server,
but it allows the collections to be completed in a shorter time.
Most of the time is spent waiting for the responses
anyway!
HTH
...Laurie :{)
OverTime: MRTG/RRDtool from within NetView
----- Original Message -----
Sent: Thursday, April 29, 2004 11:25
PM
Subject: RE: [nv-l] graph and daemon
problems
It's hard to imagine what
might be wrong since you are so far downlevel. Lots of maintenance has
been delivered since then, and as I said, I am not expert in this area.
I don't work on data collection nor graphing.
My only suggestion is that if you
have lots and lots of data then you might go into /usr/OV/app-defaults/Xnm and
increase you maxMallocPoints from the default of 5120 to something much
greater, say four times that amount.
But the real answer may lie in what happens when you
try to graph. Do you get an error message when you try to graph or
what?
Other than this, I have no
suggestions. I am hoping that some other user who does graphing all the
time will have some suggestions. Because if not, you will have to open a
problem to Support.
James
Shanks
Level 3 Support for Tivoli NetView for UNIX and
Windows
Tivoli Software / IBM Software Group
"Ioannis Yerou"
<ioannisi@ieee.org>
Sent by: owner-nv-l@lists.us.ibm.com
04/29/2004 07:31 AM
|
|
To
| <nv-l@lists.us.ibm.com>
|
|
cc
|
|
|
Subject
| RE: [nv-l] graph and
daemon problems |
|
James,
Thanks a lot
for the input. First, I checked for the ovwdb and you were right. The number
in cache was 10000 and the objects number was 12000. Therefore, I changed the
number to 15000. Now, regarding the graph I changed the collection time to 10
min but still it does not seem to work. What I have noticed in the log is
maybe what you are calling gaps. For many routers it says is not up.
Any other input regarding the graph problem will be
appreciated.
We are planning to upgrade the
version of Netview to the latest.
Thanks
again,
Regards,
Ioannis
-----Original
Message-----
From: owner-nv-l@lists.us.ibm.com
[mailto:owner-nv-l@lists.us.ibm.com] On Behalf Of James
Shanks
Sent: Wednesday, April 28, 2004 4:16 PM
To:
nv-l@lists.us.ibm.com
Subject: RE: [nv-l] graph and daemon
problems
OVs_NON_WELL_BEHAVED is not an
error condition but an indication that these daemons can stay up independently
of the core daemons like ovwdb, pmd, and trapd. The ones you cite will
always be OVs_NON_WELL_BEHAVED, as will others like nvpagerd.
You high ovwdb CPU rate on the other hand is not good. Is
your cache size properly adjusted? Do "ovobjprint -S" and see what you
object count is. Then make certain that your cache size on ovwdb is at
least 20% greater. Otherwise the daemon will spend a lot of time
allocating and reallocating storage to hold your object database in memory.
And you really should plan to move to a supported
level of the code. 7.1.1 has not been supported ever since 7.1.3 came
out over two years ago. The current maintenance level on 7.1.3 is
FixPack2, which means that you are 5 version of the code downlevel so far.
FixPack3 for 7.1.3 is in the works now.
But none of these things has anything to do with your problem.
graphing is done by xnmgraph on data collected by
snmpCollect.
I am not an expert in this area but I do
know that data will not graph if there are gaps in the data, and I'll bet that
if you look in the snmpCollect log you will find lots of errors, producing
those gaps. Collections get deferred if an error occurs, which produces
those gaps. Offhand, I would guess that you cannot successfully poll 200
routers for data in 3 minutes. If you quadruple the number of devices
to collect from, I'd suspect you have to more than quadruple the
collection interval.
Anybody
else?
James Shanks
Level
3 Support for Tivoli NetView for UNIX and Windows
Tivoli Software /
IBM Software Group
"Ioannis Yerou"
<ioannisi@ieee.org>
Sent by:
owner-nv-l@lists.us.ibm.com
04/28/2004 08:11 AM
|
|
To
| <nv-l@lists.us.ibm.com>
|
|
cc
|
|
|
Subject
| RE: [nv-l] graph and
deamon problems |
|
Hi to all,
It has been some
days now that the grapher stopped to display the BandwidthUtilIn and
BandwidthUtilOut (from MIB Data Collection) for the interfaces of the routers
selected. Even though I can see that the data is collected for all the
routers. I have changed the time interval from 1 min to 3 min but no effect. I
should also mention that I was collecting data for about 50 routers without
problems and the last month that number was increased to 200. I noticed also
the following things which I don’t know if they are related. Two deamons
trapgend and mgragentd even though running are OVs_NON_WELL_BEHAVED. The last
thing is that the process running the ovwdb takes about 99 to 100% of CPU
continuously.
Any
Ideas?
Netview 7.1.1
AIX 4.3
ARN Nortel Routers
Regards,
Ioannis
|