Mailing List Archive

Personally I like to use the snAgentCpuUtilTable, it gives the CPU
values in one hundredths of a percent, so 10000 = 100% Util. It is
available on all devices except ServerIron and the OEM edgeiron's.

me at localhost tftpboot]# snmptable SOME_FOUNDRY_SWITCH
snAgentCpuUtilTable
SNMP table: FOUNDRY-SN-AGENT-MIB::snAgentCpuUtilTable

snAgentCpuUtilSlotNum snAgentCpuUtilCpuId snAgentCpuUtilInterval
snAgentCpuUtilValue
1 1 1
1000
1 1 5
800
1 1 60
900
1 1 300
1100
2 1 1
0
2 1 5
0
2 1 60
0
2 1 300
0

What does doing an snmptable of the above table show for your device?

Here is walk of the table showing the actual OID's incase you don't have
the mib installed:

[root at emvl-nms2 archive]# snmpbulkwalk MY_SWITCH snAgentCpuUtilTable -On
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.1.1.1 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.1.1.5 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.1.1.60 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.1.1.300 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.2.1.1 = INTEGER: 2
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.2.1.5 = INTEGER: 2
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.2.1.60 = INTEGER: 2
.1.3.6.1.4.1.1991.1.1.2.11.1.1.1.2.1.300 = INTEGER: 2
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.1.1.1 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.1.1.5 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.1.1.60 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.1.1.300 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.2.1.1 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.2.1.5 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.2.1.60 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.2.2.1.300 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.1.1.1 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.1.1.5 = INTEGER: 5
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.1.1.60 = INTEGER: 60
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.1.1.300 = INTEGER: 300
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.2.1.1 = INTEGER: 1
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.2.1.5 = INTEGER: 5
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.2.1.60 = INTEGER: 60
.1.3.6.1.4.1.1991.1.1.2.11.1.1.3.2.1.300 = INTEGER: 300
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.1.1.1 = Gauge32: 700
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.1.1.5 = Gauge32: 900
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.1.1.60 = Gauge32: 1100
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.1.1.300 = Gauge32: 1100
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.2.1.1 = Gauge32: 0
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.2.1.5 = Gauge32: 0
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.2.1.60 = Gauge32: 0
.1.3.6.1.4.1.1991.1.1.2.11.1.1.4.2.1.300 = Gauge32: 0
[root at emvl-nms2 archive]#

-----Original Message-----
From: foundry-nsp-bounces@puck.nether.net
[mailto:foundry-nsp-bounces at puck.nether.net] On Behalf Of Gerald Krause
Sent: Wednesday, January 25, 2006 10:40 AM
To: foundry-nsp at puck.nether.net
Subject: [f-nsp] Problem with CPU load monitoring via SNMP

Hello folks :)

Our System: NetIron 400 + M4 + B2P8.0

I've noticed some strange difference when comparing the CPU/Process load

values coming from the local console...

N400#sh proc cpu
Process Name 5Sec(%) 1Min(%) 5Min(%) 15Min(%) Runtime(ms)
ACL 0.00 0.00 0.00 0.00 0
ARP 0.00 0.00 0.01 0.00 621932
BGP 0.02 0.03 0.03 0.03 3223299
DOT1X 0.00 0.00 0.00 0.00 0
GVRP 0.00 0.00 0.00 0.00 0
ICMP 0.46 0.55 0.51 0.46 188671
IP 7.13 8.34 7.28 6.69 21731944
IP_M 0.00 0.00 0.00 0.00 53112
IPUP 0.11 0.06 0.06 0.07 3335265
ISIS 0.00 0.00 0.00 0.00 0
L2VLAN 0.00 0.00 0.00 0.00 0
NAT 0.00 0.00 0.00 0.00 0
OSPF 0.09 0.11 0.10 0.09 6377250
RIP 0.00 0.00 0.00 0.00 253837
STP 0.00 0.00 0.00 0.00 0
VRRP 0.00 0.00 0.00 0.00 9899
IPv6 0.02 0.03 0.03 0.03 3102331
ICMP6 0.06 0.07 0.07 0.07 6094057
ND6 0.00 0.00 0.00 0.00 15139
RIPng 0.00 0.00 0.00 0.00 9502
OSPFv3 0.00 0.00 0.00 0.00 0
IPV6_RX 0.00 0.00 0.00 0.00 3

...with the ones obtained via SNMP:

admin at somehost# snmpwalk [...]
enterprises.1991.1.1.2.1.35.0 = Gauge32: 1
enterprises.1991.1.1.2.1.50.0 = Gauge32: 1
enterprises.1991.1.1.2.1.51.0 = Gauge32: 1
enterprises.1991.1.1.2.1.52.0 = Gauge32: 1

The SNMP query always produce "1" for all CPU related OID I know and not

my expected value (which would be somewhat between 7...9)!

- It's a Software bug?
- Have I referenced the wrong OID?
- Maybe the overall CPU load ist _not_ the appropriate column sum?


Thx for any hints

--Gerald
_______________________________________________
foundry-nsp mailing list
foundry-nsp at puck.nether.net
http://puck.nether.net/mailman/listinfo/foundry-nsp

On Wednesday 25 January 2006 19:54, Cliff Fogle wrote:
> Personally I like to use the snAgentCpuUtilTable, it gives the CPU
> values in one hundredths of a percent, so 10000 = 100% Util. It is
> available on all devices except ServerIron and the OEM edgeiron's.
...
> What does doing an snmptable of the above table show for your device?

The same (because 100 is here the same as 1 in the other table I used
before):

enterprises.1991.1.1.2.11.1.1.4.1.1.1 = Gauge32: 100
enterprises.1991.1.1.2.11.1.1.4.1.1.5 = Gauge32: 100
enterprises.1991.1.1.2.11.1.1.4.1.1.60 = Gauge32: 100
enterprises.1991.1.1.2.11.1.1.4.1.1.300 = Gauge32: 100

Never below, never above 100 even when I tried to stress the router a
little bit with flooding ICMPs destinated to his loopback interface.

Furthermore I found out that the "show cpu" values are equal to the SNMP
values:

# sh cpu
1 percent busy, from 121 sec ago
1 sec avg: 1 percent busy
5 sec avg: 1 percent busy
60 sec avg: 1 percent busy
300 sec avg: 1 percent busy

I guess "sh proc cpu" has nothing to do with the real CPU load value.
Maybe it represents only a more detailed process statistic related to
the "1% busy CPU"?


--Gerald

Wow. That looks like a bug.

My show proc cpu usually shows LESS cpu usage, I think because it's not
really showing all processes.

-----Original Message-----
From: Gerald Krause [mailto:gk@ax.tc]
Sent: Wednesday, January 25, 2006 12:11 PM
To: Cliff Fogle
Cc: foundry-nsp at puck.nether.net
Subject: Re: [f-nsp] Problem with CPU load monitoring via SNMP

On Wednesday 25 January 2006 19:54, Cliff Fogle wrote:
> Personally I like to use the snAgentCpuUtilTable, it gives the CPU
> values in one hundredths of a percent, so 10000 = 100% Util. It is
> available on all devices except ServerIron and the OEM edgeiron's.
...
> What does doing an snmptable of the above table show for your device?

The same (because 100 is here the same as 1 in the other table I used
before):

enterprises.1991.1.1.2.11.1.1.4.1.1.1 = Gauge32: 100
enterprises.1991.1.1.2.11.1.1.4.1.1.5 = Gauge32: 100
enterprises.1991.1.1.2.11.1.1.4.1.1.60 = Gauge32: 100
enterprises.1991.1.1.2.11.1.1.4.1.1.300 = Gauge32: 100

Never below, never above 100 even when I tried to stress the router a
little bit with flooding ICMPs destinated to his loopback interface.

Furthermore I found out that the "show cpu" values are equal to the SNMP

values:

# sh cpu
1 percent busy, from 121 sec ago
1 sec avg: 1 percent busy
5 sec avg: 1 percent busy
60 sec avg: 1 percent busy
300 sec avg: 1 percent busy

I guess "sh proc cpu" has nothing to do with the real CPU load value.
Maybe it represents only a more detailed process statistic related to
the "1% busy CPU"?


--Gerald

On Wednesday 25 January 2006 21:25, Cliff Fogle wrote:
> Wow. That looks like a bug.
>
> My show proc cpu usually shows LESS cpu usage, I think because it's
> not really showing all processes.

Hm, and my "sh proc cpu" values are constantly greater than the "sh cpu"
ones so I'am also afraid of a bug. Or I have completely missunderstood
the real relationship between "sh cpu" and "show proc cpu".


--Gerald

On Wed, 2006-01-25 at 22:12 +0100, Gerald Krause wrote:
> Hm, and my "sh proc cpu" values are constantly greater than the "sh cpu"
> ones so I'am also afraid of a bug. Or I have completely missunderstood
> the real relationship between "sh cpu" and "show proc cpu".

Did anyone ever get a real answer to this one, perhaps from a Foundry
rep..? We've got the same discrepancy and would really like to know
what's what and if we're even monitoring the right values.

Cheers,

Erik


--
---
Erik Haagsman
Network Architect
We Dare BV
Tel: +31(0)10-7507008
Fax: +31(0)10-7507005
http://www.we-dare.nl

On Wednesday 05 April 2006 18:11, Erik Haagsman wrote:
> On Wed, 2006-01-25 at 22:12 +0100, Gerald Krause wrote:
> > Hm, and my "sh proc cpu" values are constantly greater than the "sh
> > cpu" ones so I'am also afraid of a bug. Or I have completely
> > missunderstood the real relationship between "sh cpu" and "show
> > proc cpu".
>
> Did anyone ever get a real answer to this one, perhaps from a Foundry
> rep..?

Not yet :/.

--
Gerald AX/TC