SNMP and udev

[adrianmarsh]

Hi,

I'm having really confusing problems with some new SUSE 10 boxes that I've built, that my Cacti sysem is reading from (Cacti is on a windoz box)

Basically: I think the SNMP references for each machine are getting swapped around randomly on reboots (thats my running theory).

Heres a partial SNMP output before a reboot:

IF-MIB::ifDescr.1 = STRING: lo
IF-MIB::ifDescr.2 = STRING: eth1
IF-MIB::ifDescr.3 = STRING: eth0
IF-MIB::ifDescr.4 = STRING: sit0
SNMPv2-MIB::sysUpTime.0 = Timeticks: (9766778) 1 day, 3:07:47.78

And here it is again after the reboot:

IF-MIB::ifDescr.1 = STRING: lo
IF-MIB::ifDescr.2 = STRING: eth0
IF-MIB::ifDescr.3 = STRING: eth1
IF-MIB::ifDescr.4 = STRING: sit0
SNMPv2-MIB::sysUpTime.0 = Timeticks: (32994) 0:05:29.94

Which I think proves my theory right?

My graphs are set for "Uptime Goes Backwards " detection, so Cacti should detect this - right ??

I'm running 0.8.6g at the moment.

[adrianmarsh]

Actually... I just checked my SNMP cache, and after the reboot I find:

Host: RLGSIM4 - RNC339, SNMP Query: SNMP - Interface Statistics
Index: 2, Field Name: ifDescr, Field Value: eth1
OID: .1.3.6.1.2.1.2.2.1.2.2

and
Host: RLGSIM4 - RNC339, SNMP Query: SNMP - Interface Statistics
Index: 3, Field Name: ifDescr, Field Value: eth0
OID: .1.3.6.1.2.1.2.2.1.2.3

Which confirms the switch hasn't been detected.

[adrianmarsh]

Well.. I fixed it by having Cacti check everytime.. But I'm sure it should've picked up the change on the tickcounts, so I'll submit it as a bug.

[TheWitness]

As noted, the old version of Cactid would not properly re-index. When you upgrade, you will get a bunch of SQL Errors about duplicate key's. This is normal and will go away on the next poll. As fare as any other issues, I would need to see a full poll after sometime after your first.

TheWitness

[cactiabuser]

Hi,
Adrian no longer works for our company and kindly left Cacti ownership to me. We are still seeing the same problem with SNMP messages being swapped for the 2 NIC's we have installed. As such Cacti is reporting data throughputs in excess of 80Mb from the source but only a few hundred K at the destination. We are mapping the ports at both ends of a single etehernet cable and seeing this problem. We are currently running ver 0.8.6h. Is an upgrade to the latest version recommended and if so is this a simple process? I am completely new to Cacti and as such my knowledge is very limited.

[TheWitness]

Not too sure. What I can suggest is that your reindex your hosts. This typically helps these things. As an operations process, you have to pay particular attentiion when you add/remove cards from a switch, or you add interfaces in the case of a router.

When you do these things, oftentimes, your indexes will change. So, please keep that in mind. You need to understand for each device type, which field is used for indexing and what will happen when you make these changes.

TheWitness

[adrianmarsh]

hello "cactiabuser" (like I don't know you're real name...)

You need to ensure that udev hasn't re-mapped. This happens if you clone the linux machine (where the old NIC config stays in udev, and udev maps NICs to new interfaces). The SNMP agent then mismatches the configs from what you want.

Assuming you cloned one of the Linux servers, then check the config. I think I even left an old README on one of the RLGsims in the lab that explained it..


L8r..