--> af_ has joined #net-snmp
--> Haris_ has joined #net-snmp
--> jsafrane has joined #net-snmp
--> jsafrane_ has joined #net-snmp
--> ChanServ has joined #net-snmp
--> blotchy has joined #net-snmp
--> praseed has joined #net-snmp
--> zumbi has joined #net-snmp
--> darix has joined #net-snmp
--> Rada has joined #net-snmp
--> jsafrane has joined #net-snmp
--> Haris has joined #net-snmp
--> RAThomas has joined #net-snmp
--> hardaker has joined #net-snmp
--> darnok has joined #net-snmp
--> jsafrane has joined #net-snmp
--> Haris__________ has joined #net-snmp
--> Haris_ has joined #net-snmp
--> Haris_ has joined #net-snmp
--> magfr has joined #net-snmp
--> hardaker has joined #net-snmp
--> mspencer has joined #net-snmp
[2008/09/05 14:26:23] #net-snmp <praseed> Hi I'm using SNMP 5.4.1 in a powerpc based embedded system.
[2008/09/05 14:27:23] #net-snmp <praseed> Even though I'm not sending any get/set to the agent, there is always some CPU utilization by SNMP agent
[2008/09/05 14:27:56] #net-snmp <praseed> I used strace & net-snmp debug options to find why is it so
[2008/09/05 14:28:18] #net-snmp <praseed> & understood about the caching inside
--> sly has joined #net-snmp
[2008/09/05 14:29:28] #net-snmp <praseed> I removed the NETSNMP_CACHE_AUTO_RELOAD from all tables including ifTable
[2008/09/05 14:30:57] #net-snmp <praseed> However still I'm seeing lots internal get for ifTable(I don't know who is doing getnext on ifTable)
[2008/09/05 14:31:40] #net-snmp <praseed> Also I've seen constant gettimeofday function call too
[2008/09/05 14:32:16] #net-snmp <praseed> Here is the logs from net-snmp. http://pastebin.ca/1195214
[2008/09/05 14:33:04] #net-snmp <praseed> Help me bring down the CPU utilization to zero, if there is no set, get, getnext action from the management system
[2008/09/05 14:33:54] #net-snmp <rstory> what is the cpu utilization now?
[2008/09/05 14:34:07] #net-snmp <rstory> why not just exclude the ifTable?
[2008/09/05 14:42:43] #net-snmp <praseed> It ranges from 0.6 to 4 percent.
[2008/09/05 14:44:13] #net-snmp <praseed> we want ifTable support too. but only once in a while
[2008/09/05 14:45:22] #net-snmp <praseed> BTW I noticed that this 4% CPU utilzation is when there is heavy ethernet & RF traffic
[2008/09/05 14:47:19] #net-snmp <praseed> http://pastebin.ca/1195219 is the log when I started the Net-snmp agent with "strace /usr/sbin/snmpd -f -p /var/run/snmpd.pid -Dsnmpd/select,timer"
[2008/09/05 14:47:37] #net-snmp <rstory> unless the ethernet traffic is snmp traffic, net-snmp shouldn't be affecting cpu utilization.. and 0.6 when idle isn't bad. disabling auto reload will mean that we can't properly detect when the ifTable counter wrap...
[2008/09/05 14:49:21] #net-snmp <praseed> yeah I read that in the souce code comments, but I just want to make sure that's the only reason.
[2008/09/05 14:50:00] #net-snmp <praseed> The traffic is not snmp traffic, but CPU utilization goes upto 4
[2008/09/05 14:50:02] #net-snmp <praseed> %
[2008/09/05 14:50:14] #net-snmp <rstory> we also won't be able to detect interfaces going up/down, new interfaces, or deleted interfaces
[2008/09/05 14:51:25] #net-snmp <praseed> OK. is that gettimeofday logs expected?
[2008/09/05 14:52:29] #net-snmp <praseed> Also could you please have a quick look to http://pastebin.ca/1195219 which shows the startup of agent
[2008/09/05 14:52:33] #net-snmp <praseed> ?
[2008/09/05 14:52:47] #net-snmp <rstory> gettimeofday is used to calculate when to run timers and how long select should sleep.
[2008/09/05 14:56:28] #net-snmp <rstory> on my build of svn, it looks like there is a timer firing about every 5 seconds.. i think that might be for cpu stats.. disabling or increasing that time might help a little..
[2008/09/05 14:57:12] #net-snmp <praseed> OK
[2008/09/05 14:57:37] #net-snmp <praseed> I'll check CPU stats
[2008/09/05 14:57:40] #net-snmp <rstory> grep the code for snmp_alarm_register() for various alarms
[2008/09/05 14:58:44] #net-snmp <praseed> ok
[2008/09/05 15:51:51] #net-snmp <praseed> rstory: I removed the NETSNMP_CACHE_RELOAD from ifTable, why am I still seeing message
[2008/09/05 15:52:02] #net-snmp <praseed> write(3, " GETNEXT message\n", 18) = 18
[2008/09/05 15:52:02] #net-snmp <praseed> write(2, " GETNEXT message\n", 18 GETNEXT message
[2008/09/05 15:52:02] #net-snmp <praseed> ) = 18
[2008/09/05 15:52:02] #net-snmp <praseed> write(3, " -- .1.3.6.1.2.1.2.2.1.8.4\n", 30) = 30
[2008/09/05 15:52:02] #net-snmp <praseed> write(2, " -- .1.3.6.1.2.1.2.2.1.8.4\n", 30 -- .1.3.6.1.2.1.2.2.1.8.4
[2008/09/05 15:52:22] #net-snmp <praseed> Just wondering who is making this getnext call
[2008/09/05 15:52:35] #net-snmp <praseed> to ifTable
[2008/09/05 15:52:42] #net-snmp <praseed> I'm not able to trace this
[2008/09/05 15:54:25] #net-snmp <praseed> this is periodic too & I dont have a management system running
[2008/09/05 15:55:25] #net-snmp <rstory> do you have any monitors set up in snmpd.conf?
[2008/09/05 15:57:06] #net-snmp <praseed> Yes I do
[2008/09/05 15:57:29] #net-snmp <praseed> for linkUp & linkDown traps
[2008/09/05 15:57:42] #net-snmp <rstory> well, then that's what's polling ifTable
[2008/09/05 15:57:43] #net-snmp <praseed> oh u r rt
[2008/09/05 15:58:04] #net-snmp <praseed> I'm doing that for .1.3.6.1.2.1.2.2.1.8
[2008/09/05 15:58:25] #net-snmp <praseed> thanx for pointing this out
[2008/09/05 15:58:53] #net-snmp <praseed> how can I increase the poll interval
[2008/09/05 15:58:56] #net-snmp <praseed> ?
[2008/09/05 15:59:28] #net-snmp <praseed> is it monitor -r <secs> format ?
[2008/09/05 15:59:50] #net-snmp <rstory> check the snmpd.conf man page
[2008/09/05 16:01:23] #net-snmp <praseed> I got it...thanks rstory. You are my life-saver for this nth time
--> Haris_ has joined #net-snmp
--> hardaker has joined #net-snmp
--> Haris_ has joined #net-snmp
--> Lap_64 has joined #net-snmp
--> snmpbot has joined #net-snmp