Uptime plugin (reboot monitor)

mcutting · Post by **mcutting** » Fri Jun 29, 2007 1:35 am

AdrianMarsh - you need the modified host.php file that is included with the uptime plugin. However, this was a hack that was designed for PA 0.9, and doesn't work with the latest PA of 1.1.

If you are using the Manage plugin, the uptime is shown in the devices menu here.

Post by **TheWitness** » Fri Jun 29, 2007 5:13 am

I will be placing "sysUptime" in the 0.8.6k version of Cacti's host table. So, maybe that is a good thing?

TheWitness

mcutting · Post by **mcutting** » Fri Jun 29, 2007 6:51 am

Yes - I could use this as well.....

apperrault · Post by **apperrault** » Fri Jun 29, 2007 10:53 am

I don't know what is up with my system, but from some reason i cannot get this to run after the poller runs. My Windows scheduled task is as follows:

Code: Select all

e:\php\bin\php.exe e:\cacti\poller.php

and it doesn't run. Everything else is working just fine. I mean my Discovery is running every 6 hours, just like it is supposed to. The only thing that is not working is the Uptime plugin.

Does anyone have any thoughts, or should I just disable the Uptime plugin and wait for the new version later in July?

thoughts, suggestions?

app

mcutting · Post by **mcutting** » Fri Jun 29, 2007 11:29 am

App - put your poller into DEBUG for one pass, and email me the log. I'll take a look.

Regs

adrianmarsh · Post by **adrianmarsh** » Sun Jul 08, 2007 6:21 am

Anyone else seen false positives/alerts from this plugin?

Heres one, recieved email at 4am today (along with a bunch of other "host" reboot failures):

Checking into the others, one looks right but the others dont..

Host : ubiphone (Ubiphone (asterisk)), old uptime was :
60478900 (7d), new is : 29528 (0d)

But a walk of that machine shows:

C:\Program Files\Microsoft Visual Studio 8\VC>snmpwalk -c public -v1 192.168.50.6
SNMPv2-MIB::sysDescr.0 = STRING: Linux ubiphone 2.6.9-55.0.2.EL #1 Tue Jun 26 14:08:18 EDT 2007 i686
SNMPv2-MIB::sysObjectID.0 = OID: NET-SNMP-MIB::netSnmpAgentOIDs.10
DISMAN-EVENT-MIB::sysUpTimeInstance = Timeticks: (2961358) 8:13:33.58

tonzy0907 · Post by **tonzy0907** » Fri Jul 13, 2007 11:47 pm

Host : 10.*.*.* (*******), old uptime was : 665243 (0.1d), new is :
1433 (0d)

what is the meaning of 665243(0.1d) and 1433(0d)..?

sorry >? please help me ....

adrianmarsh · Post by **adrianmarsh** » Wed Aug 15, 2007 3:13 am

My thoughts: These will be the previous uptime counter before the reboot VS the current uptime counter. When the current drops below the previous, then thats what triggers the "rebooted" scenario.

However, on my system, I regularly get false positives (weekly). I'm guessing that the 32bit counter rolls over, and that the Uptime plug in can't take this into account.

adrianmarsh · Post by **adrianmarsh** » Sun Sep 16, 2007 2:22 am

Can anyone confirm my thinking on this? The false positives are getting more annoying, so I may have to turn the plugin off, even though its a great idea..

Post by **TheWitness** » Sun Sep 16, 2007 7:52 am

You should time it. I cam up with the following:

Assuming a signed integer, which it likely isn't:

248 Days till rollover.

As an unsigned integer:

497 Days till rollover. So, maybe the counter's hosed on the agent(s) in question.

Larry

adrianmarsh · Post by **adrianmarsh** » Sun Sep 16, 2007 8:08 am

Hmm... on mine its more like once a week, and its usually on several hosts at a time. Heres todays:

Host : 127.0.0.1 (ubiq-serv6 Cacti), old uptime was : 60478034 (7d), new is : 27805 (0d)

Host : ubiphone (Ubiphone (asterisk)), old uptime was :
60476756 (7d), new is : 27340 (0d)

Host : ubiq-serv3 (ubiq-serv3 SVN), old uptime was :
60474301 (7d), new is : 24564 (0d)

Al timed at 4:07am, and obviously 7d..

Post by **TheWitness** » Sun Sep 16, 2007 10:21 am

Looks like a 26bit counter. What's up with that? Or maybe they are internally incrementing timetics in micro seconds and shifting the bits.

TheWitness

adrianmarsh · Post by **adrianmarsh** » Tue Sep 18, 2007 9:50 am

Dunno.. but I get more false positives out of it than real alerts.. so unless the author can answer, I'll have to remove it..

Post by **TheWitness** » Tue Sep 18, 2007 10:06 am

I am sure the author's only response would be to contact the vendor and have them correct their agent. There is no way to fix this IMHO. There may be ways to reduce the likelihood of false positives, but to prevent it 100%. No way.

Here is an example of "predictable" behavior enforcement. If you know in advance what the 'wrap around' value is for a particular sysObjectID (aka Device Type), you could store that. And then if you measure a timetic that is within 1 poller interval of the wrap around value, set a flag to ignore the uptime going backward's if it occurs on the next poll. Then, on the next poll, reset the flag regardless.

However this is not foolproof as some vendors while maintaining the same sysObjectID will "fix" their agents between firmware/software rev.s, again making this pretty hard to lock down.

Larry

adrianmarsh · Post by **adrianmarsh** » Tue Sep 18, 2007 10:16 am

ok, but this isn't against one vendor..

Its against hosts running net-snmp Linux, against a HP printer, a Konica Printer and so on...

note: the "uptime" on these machines stays high..

Cacti

Uptime plugin (reboot monitor)

questing about uptime

Who is online