spine 500 ms timeout

jbw · Post by **jbw** » Wed Oct 29, 2008 4:01 pm

I'm suffering from the same issue. I'm trying to poll poll traffic data from only a few hosts, but these hosts have many interfaces.

If I don't use spine, things mostly work, but I do see some gaps on my graphs on a few interfaces. I thought I'd give spine a try to see if it does any better. I originally tried 0.8.7a, and today installed from SVN.

With Spine, I've only been running from the command line so far, so make sure I see the correct results. What I find is that I get SNMP timeouts when polling my two devices with lots of interfaces, but when polling another device for only 3 interfaces, it seems to show the right values.

I see from the polling setup page in the GUI that the max number of OIDs that can be polled per request is 60. Is it possible I am getting timeouts because of the high number of interfaces per host? In the output it looks like each OID is a different request.

I have attached a text file showing the output to my console when I run spine. I'd appreciate the help.

Jacek_nykis · Post by **Jacek_nykis** » Thu Jan 08, 2009 7:43 am

HI all,

I was wondering if you managed to debug this problem and find solution or workaround? I am suffering from the same timeouts and I am forced to use cmd.php poller. In my case if I run spine for multiple hosts first one is polled fine all the rest fails. If poll one host at the time everything works as expected.

Post by **gandalf** » Fri Jan 09, 2009 10:06 am

Please make sure to use latest spine, e.g. 087c. There have been quite a lot of bug fixes. Do NOT use it in production yet!
Reinhard

Jacek_nykis · Post by **Jacek_nykis** » Sat Jan 10, 2009 7:52 am

Hi,

I have just compiled latest SVN version and gave it a try. This did not make any difference, I still have timeouts.

spine --version
SPINE 0.8.7c-beta3 Copyright 2002-2008 by The Cacti Group

I do not know if it matters but stats seem wrong - they always show that number of hosts is one more that actual number. If I use spine for one host only I get:
SPINE: Poller[0] Time: 0.5695 s, Threads: 1, Hosts: 2

Regards
Jacek

Post by **gandalf** » Thu Jan 15, 2009 2:28 pm

I fear this is for Larry (TheWitness) only
Reinhard

Jacek_nykis · Post by **Jacek_nykis** » Thu Jan 29, 2009 9:54 am

gandalf wrote:I fear this is for Larry (TheWitness) only
Reinhard

Hi,

Thank you Reinhard.
This behavior is persistent in my environment. I have just tried spine 0.8.7c from SVN and it made no difference.
Would you like me to submit a bug?

Regards
Jacek

Post by **gandalf** » Sat Jan 31, 2009 6:33 am

A bug or a pm to TheWitness
Reinhard

Maestro_Ba · Post by **Maestro_Ba** » Thu Mar 12, 2009 9:14 am

Hi there. I'm experiencing the same problem, I'm getting "SNMP timeout detected" messages, and changing the value in Settings->General doesn't do anything.

Has this been fixed in the latest spine release (0.8.7c)? I do not want to stop my production machine until there is a stable release that corrects this...

Thanks a lot.

Post by **TheWitness** » Thu Mar 12, 2009 4:50 pm

You must adjust the MAX OID's to something your device can handle, as little as 1. The quality of the SNMP agent and the strength of the CPU dictate how high this number can go.

TheWitness

Post by **TheWitness** » Thu Mar 12, 2009 4:50 pm

This is a per host setting in 0.8.7c/d.

TheWitness

spine 500 ms timeout

Who is online