ProCurve 2524: monitor traffic usage by port

[zwar]

Hello, i'd like to monitor how many (giga)bytes are transfered over a specific port on the switch. I know the web, cli and telnet-menu interfaces offer you port traffic summaries. But based on the current numbers I have some doubts wheather these are accurate or i simply do not understand something.

For example: one of the ports shows me there have been 2,465,389,780 bytes transmitted (Tx) over this port. But based on other data i'm sure there has been way much more traffic than 2.2 GB on that port. Is there a maximum number on which the switch begins to count from 0? Or am i missing something else?

The switch is in use for some years now and has not been reset for a long time.

[Steve Britt · Accepted solution]

Zwar,

If the web, CLI, or telnet-menu counters are based on 32-bit internal counters then they will reach a maximum value of 4,294,967,295 and roll back over to 0. For octets, 4.2GB is not a very large number. At a steady rate of 10 Mbps (that's bits/second) a 32-bit octet counter would roll in 3436 seconds (about 57 minutes); for a steady flow of 1 Gbps this number becomes 344 seconds (less than 6 minutes).

That being said, I don't know offhand whether these UI counters are based on 32-bit or 64-bit counters. I kind of suspect that only 32-bit counters are used because with 64-bit counters you should see much larger numbers on some of your ports than the number you cite if the switch has been in use and not reset for a long time. But you can eliminate any uncertainty in this regard by using SNMP to pull the counts rather than relying on the supplied UIs.

SNMP would allow you to select which size of counter you consult as the 2524 supports both 32-bit (from the interface MIB - iso(1).org(3).dod(6).internet(1).mgmt(2).mib-2(1).interfaces(2).ifTable(2)) and 64-bit (from the interface extensions MIB - iso(1).org(3).dod(6).internet(1).mgmt(2).mib-2(1).ifMIB(31).ifMIBObjects(1).ifXTable(1)) counters. Using SNMP is the path I'd recommend for your situation ...

Regards,

SVB

[Mohammed Faiz]

Hi,

If you do a "show interface X hc", where X is the number you are insterested in, it will show you the traffic counters in hex format.

HTH

[zwar]

Steve: Thank you for pointing me into the right direction, i'll try out SNMP!

Mohammed: thank you for the suggestion, but the counter turns around at the max value and displaying it in hex won't help.

[Mohammed Faiz]

Hi,

The suggestion that the counters roll over sounds reasonable but I'm not sure it's accurate.
For example, taking the output of a "sh int" on one of the interfaces of a switch gives me:

Name :
Link Status : Up
Totals (Since boot or last clear) :
Bytes Rx : 3,025,718,262
Bytes Tx : 1,559,636,547

A "show int hc" for the same port gives me:

Name :
Link Status : Up
Bytes Rx : 0x38fb48e3d0a
Bytes Tx : 0x3205d10f5cc

Converted to decimal those two numbers differ significantly.
Either way, SNMP is still the best way to achieve what you want (I can recommend Cacti).

[zwar]

Thank you for additional explanation.

I was digging in the 2425 documentation on how to setup snmp traps for this kind of information (port traffic counters), i defined a community and a trap receiver (on the receiver's end there is a service listening and logging snmp traps), but i don't have any paper covering trap definitions, how to get use of all the mibs (like e.g. stat.mib in my case)? The only thing i got is to use rmon in hp toptools. I don't want to use toptools, is there any other way?

[Mohammed Faiz]

I don't believe there is a way of configuring the switch to send interface statistics as traps.
You need something that will poll the switch for statistics (using SNMP) at a regular interval, like Cacti.

[zwar]

Thank you for that, I'll try Cacti.