Unsolved
This post is more than 5 years old
6 Posts
0
8316
December 3rd, 2003 16:00
bug in 64 bit SNMP counters on 5224
We've had about a dozen 5224 switches for about 6 months now and we've never been able to correctly grab 64bit interface counters.
So, I upgraded to the latest firmware (3.1.0.8) and unfortunately SNMP is still broken.
I made a little perl script that grabs the OID and shows em every second. The 2nd number is a timestamp.
296281940524 (1070473595)
296298160685 (1070473596)
296312559575 (1070473597)
296312559575 (1070473598)
296328633236 (1070473600)
296345089215 (1070473601)
300653909877 (1070473602)
300653909877 (1070473603)
300670727487 (1070473604)
300685553370 (1070473605)
300700547137 (1070473606)
And a bit later:
304910771356 (1070474021)
304925600919 (1070474022)
304939320230 (1070474023)
304939320230 (1070474025)
309249303692 (1070474026)
309263450505 (1070474027)
309278564509 (1070474028)
309293106454 (1070474029)
309293106454 (1070474030)
As you can see, it suddenly increases by a huge number in a 1 second period. This causes annoying peaks in graphs and average
counters to be way off. It seems the increase is exactly 4Gbyte, so it is probably some bit that is flipped the wrong way.
Could this please be fixed sometime because there actually are people who think SNMP is important :)
Regards,
Cor
So, I upgraded to the latest firmware (3.1.0.8) and unfortunately SNMP is still broken.
I made a little perl script that grabs the OID and shows em every second. The 2nd number is a timestamp.
296281940524 (1070473595)
296298160685 (1070473596)
296312559575 (1070473597)
296312559575 (1070473598)
296328633236 (1070473600)
296345089215 (1070473601)
300653909877 (1070473602)
300653909877 (1070473603)
300670727487 (1070473604)
300685553370 (1070473605)
300700547137 (1070473606)
And a bit later:
304910771356 (1070474021)
304925600919 (1070474022)
304939320230 (1070474023)
304939320230 (1070474025)
309249303692 (1070474026)
309263450505 (1070474027)
309278564509 (1070474028)
309293106454 (1070474029)
309293106454 (1070474030)
As you can see, it suddenly increases by a huge number in a 1 second period. This causes annoying peaks in graphs and average
counters to be way off. It seems the increase is exactly 4Gbyte, so it is probably some bit that is flipped the wrong way.
Could this please be fixed sometime because there actually are people who think SNMP is important :)
Regards,
Cor
0 events found
No Events found!


chiparus
1 Message
0
February 16th, 2004 13:00
What is DELL going to do about it? Switch "management" is pretty useless if you can't use 64 bit counters. What if other bugs are to be found? Do we have to wait a few years before DELL starts to fix it?
agrelic
4 Posts
0
August 18th, 2004 01:00
What's this mean? Lets look at the 64-bit counter as two 32-bit counters. At 4GB, the 'low' int is set to all 0's, and the 'high' int has its low bit set to 1, and the other bits set to zero. So what it seems is happening is when the lower int is getting rolled over, the higher int is getting incremented twice instead of once. Hopefully with this data, you can get your snmp tools to fix for it. Although in reality I wish they'd just fix the firmware.
Another way of looking at it is that the higher int should be divided by two, or bitshifted right by one. Unfortunatley this bug also means that the 64-bit counter will rollover sooner than it should :(
ThiasDude
23 Posts
0
May 24th, 2005 15:00
Matthias
chrisbyrd
1 Message
0
December 1st, 2006 08:00
We've just bought a few of these and they're kinda useless if we cant graph the gig ports properly :(
ThiasDude
23 Posts
0
December 21st, 2006 15:00