Archived community.zenoss.org | full text search
Skip navigation
Currently Being Moderated

Dev chat 05/28/2009

VERSION 1 
Created on: Sep 14, 2009 11:17 AM by Noel Brockett - Last Modified:  Sep 14, 2009 11:17 AM by Noel Brockett
[2009-05-28 10::01:52] mrayzenoss: Greetings everyone, JP and I will be answering any questions you have
[2009-05-28 10::01:58] ** jplouis joined the chat room.
[2009-05-28 10::02:00] ** jmp242 joined the chat room.
[2009-05-28 10::02:08] ** You changed the topic to "Zenoss developers are here to answer your questions".
[2009-05-28 10::02:32] Weetos: genji0714> this may mean that zenoss doesn't care what's being answered
[2009-05-28 10::03:11] ** __ian___ joined the chat room.
[2009-05-28 10::03:14] genji0714: Weetos>?? Zenoss is ignoring what the switches are responding?
[2009-05-28 10::03:30] Weetos: genji0714> it seems ... don't know why
[2009-05-28 10::03:53] Dieterbe: Hi, when I add a Server/Linux and have snmp working (I have cpu, memory etc graphs), shouldn't i also get network graphs for the interfaces? i tried going to the device->os->interfaces, select them and do "enable monitoring" but this does not give me graphs
[2009-05-28 10::03:55] genji0714: Weetos> Yea....arghh zenoss zenoss zenoss....
[2009-05-28 10::04:07] Dieterbe: @ mrayzenoss ^^
[2009-05-28 10::04:40] Jane_Curry: Dieterbe - yup - you should see each interface under the os tab - shouldn't need to activate monitoring
[2009-05-28 10::04:44] ** mrchippy joined the chat room.
[2009-05-28 10::04:51] genji0714: npmccallum_>About half of my switches are reporting that the SNMP agent is down. when in fact it is not...i am able to do an snmpwalk from Zenoss and the other half of the switches that have the exact same config..some even in the same net closet work just fine
[2009-05-28 10::05:02] mrayzenoss: Dieterbe: Debian 5 right?
[2009-05-28 10::05:03] Dieterbe: Jane_Curry: but i want graphs
[2009-05-28 10::06:06] npmccallum_: genji0714: you probably don't have the right snmp community or snmp version
[2009-05-28 10::06:19] mrayzenoss: Dieterbe: yes you should be getting graphs for your network interfaces
[2009-05-28 10::06:22] Jane_Curry: Dieterbe - you should have em - is your snmp agent configured ok? If you do an snmpwalk of the interfaces table for the device, do you get the right data?
[2009-05-28 10::06:34] genji0714: npmccallum> I have checked the comm and snmp version
[2009-05-28 10::06:34] Younger: mrayzenoss and JP: i have a router with snmp version set to v2c. so ethernetCsmacd_64 should be used instead of ethernetCsmacd. but the interface throughput graph is still drawn using ethernetCsmacd. Does any one know how to force ethernetCsmacd_64?
[2009-05-28 10::07:52] mrayzenoss: Younger: you have the ethernetCsmacd_64 template bound?
[2009-05-28 10::07:52] npmccallum_: genji0714: what happens when you go into the templates for the device, open an snmp data source and run "test" against one of the switches
[2009-05-28 10::08:04] genji0714: npmmccallum>I am using snmpv3..and i have the exact same switch model...same IOS in the same closet that works just fine. I am able to do a snmpwalk from zenoss and it works fine on the same switch that is reporting that snmp agent is down. I have tried debugging zenperfsnmp with no luck...I am done a tcpdump to no avail...i just don't know where to turn now.
[2009-05-28 10::08:44] ** cgibbons joined the chat room.
[2009-05-28 10::08:44] ** cgibbons was promoted to operator by ChanServ.
[2009-05-28 10::09:18] Jane_Curry: Dieterbe - another thought. When you click on an interface, what do you actually see? Nothing? Graph templates but no lines or data? Or graph otlines with values of "nan"?
[2009-05-28 10::09:30] Dieterbe: mrayzenoss / Jane_Curry , oh i just have to click on the interface name to see the graphs. i thought they would show up under "Perf" . is there a way to just see all bytes/sec graphs for all interfaces (you know, one image with multiple graphs)
[2009-05-28 10::09:39] Younger: mrayzenoss: By default, both ethernetCsmacd and ethernetCsmacd_64 are bound to /Devices.
[2009-05-28 10::11:08] genji0714: npmmccallum>this is what I get....Executing command snmpwalk -c -v3 10.12.241.13 1.3.6.1.2.1.2.2.1.10 against ad-a-2960-13.admin.rhnet.org Timeout: No Response from 10.12.241.13 DONE in 6 seconds
[2009-05-28 10::11:29] ** ian_ joined the chat room.
[2009-05-28 10::11:31] Jane_Curry: Dieterbe - sounds like you have found the graphs then??? The interface graphs are "special" in that their templates are bound automatically based on the name of your interface (like ethernetCsmacd)
[2009-05-28 10::14:32] ian_: Relatively so, yes.
[2009-05-28 10::14:51] jplouis: Dieterbe - you might get the graph you want by creating a custom Multi-Graph report
[2009-05-28 10::15:08] genji0714: npmmccallum>I acutally ran the same thing against one of the switches that is gathering snmp data and it timed out also....
[2009-05-28 10::16:27] Jane_Curry: ian_ - days or a week or two (without any implied committment...)
[2009-05-28 10::16:59] jplouis: not days
[2009-05-28 10::17:57] Jane_Curry: Performance threshold question - I want a threshold based on an expression of SNMP MIB variables
[2009-05-28 10::17:58] mrayzenoss: Jane_Curry: you can watch progress on 2.4.2 here: http://dev.zenoss.org/trac/report/6
[2009-05-28 10::18:19] ian_: Yeah, as long as no commitment, hopefully by end of next week? I think.
[2009-05-28 10::18:40] mrayzenoss: you did
[2009-05-28 10::19:02] ian_: I try to put things together from the bits I glean during meetings, but rarely do I get a fair representation of reality.
[2009-05-28 10::19:06] npmccallum_: genji0714: there is a syntax error in that command, considering that the '-c' is empty
[2009-05-28 10::19:38] Dieterbe: thanks Jane_Curry and jplouis . what i basically want is a centralized overview of the "perf" properties, but also the bytes/sec graphs for the interfaces, so i can easily compare the amount of network traffic to cpu load for example
[2009-05-28 10::19:49] genji0714: npmmcallum: is it something I have done...or is it a bug?
[2009-05-28 10::20:02] mrayzenoss: Jane_Curry: what thresholds do you need? I've gotten several new ZenPacks for custom thresholds from rocket
[2009-05-28 10::20:05] Jane_Curry: I can create a graph based on SNMP expressions by using my own graph command but that doesn't get me an actual datapoint and I think I can only threshold on datapoints?
[2009-05-28 10::20:42] ** NoFX_SBC joined the chat room.
[2009-05-28 10::21:47] rocket: Jane_Curry: I believe at the moment you can only threshold on datapoints as you have indicated ..
[2009-05-28 10::21:59] ** __ian___ left the chat room.
[2009-05-28 10::22:02] Weetos: mrayzenoss> the broken map bug that appeared in 2.4.0 (see http://forums.zenoss.com/viewtopic.php?t=9714) isn't fixed in 2.4.1 and I can't seem to find a ticket about that issue - Does that mean we're only a few users experiencing that problem ?
[2009-05-28 10::22:10] Jane_Curry: probaly simple SNMP expressions like oid1 - oid2 (or even oid1 oid2 - if I have to do it in Reverse Polish!)
[2009-05-28 10::25:04] rocket: Jane_Curry: not an expression .. but I have a threshold that creates custom rrd files
[2009-05-28 10::26:23] Jane_Curry: rocket - not with you there? You have a (single oid) threshold that create an rrd datafile??
[2009-05-28 10::27:26] rocket: Jane_Curry: yes .. I have submitted it to matt .. but since it replicates some enterprise functionality I dont really want to elaborate too much until they announce it
[2009-05-28 10::27:39] Jane_Curry: ok
[2009-05-28 10::28:05] tyler79: does anyone know how to deal with the issue where zentrap core dums on freebsd?
[2009-05-28 10::28:09] Jane_Curry: Matt - is there anything in the pipeline that might address thresholding on an expression??
[2009-05-28 10::28:14] rocket: its possible to do what you want from a threshold .. but its sort of a hacked way to do it
[2009-05-28 10::29:38] genji0714: npmccallum_: do you have any suggestions of how I could futher troubleshoot this issue?
[2009-05-28 10::29:44] rocket: Jane_Curry: otherwise if you do it per threshold you have to make a new threshold type for every custom threshold that comes out
[2009-05-28 10::30:10] mrayzenoss: Jane_Curry: nothing from Zenoss development that I'm aware of for new threshold types
[2009-05-28 10::30:10] npmccallum_: ian_: any suggestions for genji0714 ?
[2009-05-28 10::30:30] Jane_Curry: Yup - with you - I guess it depends on Zenoss delivering (which doesn't sound scheduled)
[2009-05-28 10::30:31] mrayzenoss: Jane_Curry: but I'll be pushing people to use rocket's thresholds more once I get them all out
[2009-05-28 10::31:12] Jane_Curry: matt - are rocket's ZenPacks available now or still in the pipeline??
[2009-05-28 10::31:21] mrayzenoss: his Point Threshold is out
[2009-05-28 10::31:42] rocket: all thresholds only work with 2.4
[2009-05-28 10::31:49] Jane_Curry: ok - got that - what ZenPack am I looking for
[2009-05-28 10::31:53] mrayzenoss: I've just been swamped so far, good response on the ZenPack contest so far
[2009-05-28 10::31:56] rocket: Jane_Curry: predictive is the one I was referring to
[2009-05-28 10::32:02] ian_: npmccallum_: Actually I missed most of the conversation, as I lost power. So, no.
[2009-05-28 10::32:10] ** Da-Geek left the chat room.
[2009-05-28 10::32:20] rocket: Jane_Curry: it creates the predictive rrd for use in that threshold
[2009-05-28 10::32:22] npmccallum_: ian_: its mostly about snmpv3, which I have no experience with
[2009-05-28 10::32:32] ian_: npmccallum_: me either.
[2009-05-28 10::32:38] npmccallum_:
[2009-05-28 10::34:31] genji0714: npmmccallum_: it is the weirdest thing! If it didn't work on anything it wouldn't be so frustrating...but because it works on some of the switches it is driving me nuts! I appreciate your help though!
[2009-05-28 10::34:56] Jane_Curry: Is this the "PointThreshold" ZenPack that we are referring to??
[2009-05-28 10::35:30] rocket: Jane_Curry: the Point threshold sets off an alert at a value
[2009-05-28 10::35:30] jplouis: I didn't quite understand the snmpv3 problem. A ticket or forum post may be easier to follow.
[2009-05-28 10::35:45] rocket: Jane_Curry: eg 4.0 just that single value
[2009-05-28 10::35:59] genji0714: jplous: I do have a forum post...not many hits
[2009-05-28 10::36:11] Jane_Curry: rocket - is this the correct ZenPack though that you and Matt have been referring to?
[2009-05-28 10::36:28] rocket: Jane_Curry: Predictive threshold .. in the pipeline yet
[2009-05-28 10::36:46] jplouis: genji0714 - link?
[2009-05-28 10::37:06] genji0714: where is the 'test' on the data sources pulling in the snmpwalk command info from?
[2009-05-28 10::37:31] Jane_Curry: ok - gotcha - I'll look out for the predictive threshold one
[2009-05-28 10::37:32] genji0714: jploous:http://forums.zenoss.com/viewtopic.php?t=9884
[2009-05-28 10::41:45] npmccallum_: genji0714: in my experience, "works sometimes" == configuration issue
[2009-05-28 10::42:56] genji0714: npmmccallum_: yea...I went down that road..I made sure that Zenoss was configured right...i have the zproperties set on the switch container so it affects all the switches...i then reconfigured one of the troublesome switches to no avail....
[2009-05-28 10::43:25] npmccallum_: genji0714: are you sure the snmp settings aren't overridden on the devices themselves?
[2009-05-28 10::44:37] genji0714: npmccallum_: overridden??
[2009-05-28 10::45:14] Jane_Curry: genji0714 - is there anything in the event console or the logs? If the problem is with the OID instance, this often shows up in the event console as a debud severity (ie normally hidden)
[2009-05-28 10::45:29] npmccallum_: genji0714: navigate to the device, from the page menu choose More -> zProperties
[2009-05-28 10::45:51] genji0714: OK
[2009-05-28 10::48:01] jplouis: genji0714 - Are you asking for the test command code that generates the snmpwalk on the datasource?
[2009-05-28 10::48:13] npmccallum_: genji0714: check your snmp community, etc
[2009-05-28 10::48:25] genji0714: npmccallum_: yea the settings are correct...
[2009-05-28 10::48:44] npmccallum_: genji0714: idk then
[2009-05-28 10::49:09] genji0714: jplouis: the test command...where does it pull its info to decided with command switches to use ---> Ie. -c -v3?
[2009-05-28 10::49:33] jplouis: genji0714 - it is in python code, not configurable
[2009-05-28 10::49:44] genji0714: jplouis: I dunno....I am really grasping for straws at this point...
[2009-05-28 10::50:04] jplouis: genji0714 - it is in BasicDataSource.py
[2009-05-28 10::50:19] genji0714: jplouis: OK, thanks
[2009-05-28 10::50:26] jplouis: "snmpwalk -c%(zSnmpCommunity)s "
[2009-05-28 10::51:04] Jane_Curry: genji0714 - anything in zenhub.log or events.log?
[2009-05-28 10::51:13] npmccallum_: genji0714: you'll notice that in the "test against device" results you posted above, the '-c' was *not* followed by a community string
[2009-05-28 10::51:24] genji0714: jplouis: oh...here is another weird thing...when I model the switch that reports 'snmp agent down' it pulls in snmp data such as the snpm desc.....
[2009-05-28 10::51:39] npmccallum_: genji0714: that would be my guess, that zSnmpCommunity is not being filled in for some reason
[2009-05-28 10::51:45] genji0714: jplouis:yea....which makes sense I have no snmp comm strings because I am using snmpv3
[2009-05-28 10::52:32] npmccallum_: genji0714: ok, I unfortunately know little about snmpv3...
[2009-05-28 10::53:04] jplouis: genji0714 - did you try an snmp get like the forum suggested? The actual collection uses get
[2009-05-28 10::53:05] genji0714: npmccallum_> it is fully supported though correct?
[2009-05-28 10::53:13] Jane_Curry: But there are those customers who insist on it - it's not that hard when you've done one!
[2009-05-28 10::53:33] genji0714: jplouis> I have not yet done an snmpget....
[2009-05-28 10::53:35] Jane_Curry: I have used SNMP V3 successfully from Zenoss to net-snmp agents
[2009-05-28 10::53:39] npmccallum_: genji0714: yes, it is fully supported, you're just asking the wrong guy
[2009-05-28 10::54:03] genji0714: npmccallum_:
[2009-05-28 10::56:06] jmp242: not really related to SNMPv3, but I have occasionally on an overloaded Windows desktop box seen SNMP down alerts in Zenoss, but they go away when the load eases up.
[2009-05-28 10::56:26] ** kevin7kal left the chat room.
[2009-05-28 10::56:50] genji0714: jmp242: no.
[2009-05-28 10::57:34] jmp242: do the switch logs (if there are any) on the failing switches show anything useful
[2009-05-28 10::57:42] genji0714: no
[2009-05-28 10::59:06] Jane_Curry: ... and the traffic arrives at the switch??? Didi the TCPDUMP on your Zenoss system show a response coming back?
[2009-05-28 10::59:36] genji0714: yrea
[2009-05-28 11::00:31] jmp242: you're not running any firewall on the zenoss box are you?
[2009-05-28 11::00:34] Jane_Curry: Can you configure a switch so that it DOES send TRAPs (V1 or V2) to the Zenoss system - if the problem is authentication then you should get an authentication TRAP
[2009-05-28 11::01:45] tyler79: zentrap is core dumping for me on freebsd 7.2. Here's the backtrace: http://pastebin.com/d6f53c6a9
[2009-05-28 11::01:47] ** kisielk joined the chat room.
[2009-05-28 11::01:51] tyler79: can anyone take a look?
[2009-05-28 11::02:01] genji0714: Jane_Curry: I will try that
[2009-05-28 11::07:18] tyler79: actually, I get a similar core dump when I try to add a device: http://pastebin.com/m3abbf5c9
[2009-05-28 11::08:30] ** kevin7kal joined the chat room.
[2009-05-28 11::09:50] ** Serideru joined the chat room.
[2009-05-28 11::10:16] ** kevin7kal left the chat room.
[2009-05-28 11::10:35] mrayzenoss: tyler79: I'm not sure what the issue is, perhaps some of the other FreeBSD users might chime in? perr0, caseyh or sergey if you see them
[2009-05-28 11::12:11] Jane_Curry: tyler79 - sorry - looked at you dumps - no idea
[2009-05-28 11::14:08] ** Younger left the chat room.
[2009-05-28 11::14:11] Jane_Curry: .. how do I cope with mutiple data tables that have DIFFERENT instance indexes?
[2009-05-28 11::26:17] npmccallum_: Jane_Curry: probably the easiest way is to write a zencommand
[2009-05-28 11::30:38] ** kevin7kal joined the chat room.
[2009-05-28 11::32:01] ** jplouis left the chat room.
[2009-05-28 11::32:10] ** ganiman joined the chat room.
[2009-05-28 11::32:20] ganiman: hello
[2009-05-28 11::33:26] venturaville: anyone have a really large infrastructure that might have some tuning tips?
[2009-05-28 11::34:27] ganiman: what is large? lol
[2009-05-28 11::34:58] venturaville: thousands of devices, and 100Ks of data points
[2009-05-28 11::35:09] ganiman: we use zenoss to monitor 2500+ devices
[2009-05-28 11::35:09] npmccallum_: venturaville: "tuning tips" don't really mean anything
[2009-05-28 11::35:19] mrayzenoss: there are some tips on these threads: http://forums.zenoss.com/viewtopic.php?t=6230 http://forums.zenoss.com/viewtopic.php?p=21843
[2009-05-28 11::35:52] ganiman: we use many boxes
[2009-05-28 11::36:07] npmccallum_: venturaville: what problems are you having?
[2009-05-28 11::36:25] ** Joao joined the chat room.
[2009-05-28 11::36:26] npmccallum_: (if you aren't having any, you probably don't need any tuning advice)
[2009-05-28 11::36:29] venturaville: not entirely certain, although zenhub is eating up most of the CPU time
[2009-05-28 11::36:37] ** Joao is now known as JoaoS.
[2009-05-28 11::36:37] venturaville: have 3 workers running now
[2009-05-28 11::37:41] npmccallum_: what is your load?
[2009-05-28 11::38:49] ganiman: can someone help me with an rrdtool graph command? I'd like to do something similar to how zenoss draws the threshold lines on graphs. If the threshold line has a far greater value than other lines being drawn, I don't want the threshold line to stretch my graph out - I'd rather just not draw it, and this is the behavior graphs generated in zenoss have.
[2009-05-28 11::39:21] venturaville: load average is not that informative (it has 4 cores)
[2009-05-28 11::40:20] npmccallum_: is zenhub using all your CPU or is it mostly IO Wait?
[2009-05-28 11::40:22] venturaville: snmp collection time has climbed to about 140 seconds on average and spikes around 4 minutes
[2009-05-28 11::41:20] npmccallum_: not cpu wait, iowait
[2009-05-28 11::41:28] venturaville: was getting to that :-)
[2009-05-28 11::41:32] ** mrchippy left the chat room.
[2009-05-28 11::41:33] venturaville: had to wait on iostat to run
[2009-05-28 11::41:50] npmccallum_: what does top say?
[2009-05-28 11::41:56] venturaville: await is about 10.63 on the data disks
[2009-05-28 11::43:02] npmccallum_: ok, so it is likely not a disk bottleneck
[2009-05-28 11::43:25] venturaville: doesn't look it, though I suppose it could be spiking in there somewhere
[2009-05-28 11::48:59] mrayzenoss: Thanks for showing up everyone, I'll have the transcript up later this afternoon
[2009-05-28 11::49:10] ** You changed the topic to "Vote for Zenoss in the SourceForge Community Choice Award Nominations: http://tinyurl.com/oyut4c".
[2009-05-28 11::49:24] kisielk: question about tcp/ip service monitoring. For some services like rsync and SSH, I end up with a whole bunch of crap in the syslog. eg "Did not receive identification string from 10.3.0.68"
[2009-05-28 11::50:01] mrayzenoss: kisielk: map the events to get dropped
[2009-05-28 11::51:05] kisielk: yeah, it's not a matter of mapping it in Zenoss
[2009-05-28 11::51:34] mrayzenoss: ahhh
[2009-05-28 11::51:37] kisielk: in Zenoss I already drop the events
[2009-05-28 11::51:45] ** JoaoS left the chat room.
[2009-05-28 11::54:13] mrayzenoss: kisielk: I don't really have a good answer for that
[2009-05-28 11::55:03] rocket: kisielk: using syslog-ng?
[2009-05-28 11::55:13] kisielk: rocket: yes
[2009-05-28 11::55:33] rocket: kisielk: create a filter in syslog-ng to send those to /dev/null first?
[2009-05-28 11::56:05] kisielk: yeah I thought of that, but I don't like the idea of regexp filtering messages on all my servers
[2009-05-28 11::56:33] rocket: hrmm I see .. I assumed you were doing central syslogging
[2009-05-28 12::02:39] kisielk: well, we do that too
Comments (0)