Long uptime cause problem with caunters

[Gorbachov]

Hello friends,

I have several machines that soon will reach 500 days uptime.

Today for two of them was 497-th day of uptime. The machines are up and running but the graphs look bad. Take a look at the attachent.


The output of the SNMP request looks okey but the numbers that came from the servers look oddly small...

Code: Select all

07/22/2009 06:01:00 PM - CMDPHP: Poller[0] Host[13] DS[487] SNMP: v2: 192.168.2.21, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 1685054545
07/22/2009 06:00:52 PM - CMDPHP: Poller[0] Host[1] DS[494] SNMP: v2: 127.0.0.1, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 538341423
07/22/2009 06:00:20 PM - CMDPHP: Poller[0] Host[19] DS[489] SNMP: v2: 192.168.2.122, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 970367048
07/22/2009 06:00:18 PM - CMDPHP: Poller[0] Host[21] DS[491] SNMP: v2: 192.168.2.142, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 3336426087
[b]07/22/2009 06:00:14 PM - CMDPHP: Poller[0] Host[17] DS[493] SNMP: v2: 192.168.2.102, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 224553
07/22/2009 06:00:06 PM - CMDPHP: Poller[0] Host[16] DS[492] SNMP: v2: 192.168.2.101, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 323519[/b]
07/22/2009 06:00:05 PM - CMDPHP: Poller[0] Host[31] DS[789] SNMP: v2: 192.168.2.2, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 3069520555
07/22/2009 06:00:04 PM - CMDPHP: Poller[0] Host[18] DS[488] SNMP: v2: 192.168.2.121, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 2442037680
07/22/2009 06:00:04 PM - CMDPHP: Poller[0] Host[20] DS[490] SNMP: v2: 192.168.2.141, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 3285588858
07/22/2009 06:00:04 PM - CMDPHP: Poller[0] Host[14] DS[495] SNMP: v2: 192.168.2.22, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 2011504797
07/22/2009 06:00:02 PM - CMDPHP: Poller[0] Host[36] DS[1051] SNMP: v1: 192.168.1.40, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 262094333

07/22/2009 06:00:14 PM - CMDPHP: Poller[0] Host[17] DS[493] SNMP: v2: 192.168.2.102, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 224553
07/22/2009 06:00:06 PM - CMDPHP: Poller[0] Host[16] DS[492] SNMP: v2: 192.168.2.101, dsname: ucd_hrSystemUptime, oid: .1.3.6.1.2.1.25.1.1.0, output: 323519

This rows are the two servers. It looks like their counters "Restart".


Is this something connected to 32/64 bit counters? Or where is the problem...


Thanks in advance
Yasen

[Gorbachov]

After small research it looks like, the counter of linux boxes reset but on old kernels? I have new kenel on red hat 5.1 and the console shows real uptime:

[root@ob-web1 ~]# uptime
15:39:05 up 497 days, 4:00, 1 user, load average: 0.12, 0.08, 0.08

[gandalf]

There are multiple "uptimes". One is the server uptime, the other one I'm aware of is the uptime of the snmpd. On many distro's, the snmpd still will be restarted regularly, so both times will differ
Reinhard

[Gorbachov]

Hello Gandalf,

It looks like the problem is with the SNMPd do you know a way to resolve this "problem"?

I will dig if there is a version of SNMPd that can bare such big uptime.

[schurzi]

Hi Gorbachov,

I think this is a problem of 32-bit counters.

You are already using the right SNMP Value for getting the uptime. (The one which resets everytime SNMPd restarts would be .1.3.6.1.2.1.1.3.0 - sysUpTime)

If you do the maths, you will see that a 32-bit counter will reset in roughly 496 Days.

Timeticks is defined in the MIB "SNMPv2-SMI.txt"

-- hundredths of seconds since an epoch
TimeTicks ::=
[APPLICATION 3]
IMPLICIT INTEGER (0..4294967295)

Code: Select all

# echo "4292967295/100/60/60/24" |  bc
496

I don't know a SNMP counter that counts uptime in 64-bit.

[Gorbachov]

Thanks schurzi,


It looks like I have to live with this