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

Dev Chat 11/11/10

VERSION 1 
Created on: Dec 13, 2010 4:51 PM by Nick Yeates - Last Modified:  Dec 13, 2010 4:55 PM by Nick Yeates
[11-Nov-2010 10:38:48] <nyeates> Hola all
[11-Nov-2010 10:39:12] <rmatte> hola
[11-Nov-2010 10:39:30] <nyeates> today we will run our gathering in spanish
[11-Nov-2010 10:39:34] <nyeates> ...kidding :-)
[11-Nov-2010 10:39:35] <rmatte> haha
[11-Nov-2010 10:39:50] <rmatte> forgot that it's developer day
[11-Nov-2010 10:40:01] <nyeates> 20 mins
[11-Nov-2010 10:40:35] <rmatte> Simon4: I assumed that's what the ACL views were for
[11-Nov-2010 10:40:57] <rmatte> I've never actually seen them in action
[11-Nov-2010 10:41:08] <Simon4> rmatte: I haven't seen anything obvious that says that, but I haven't dug too far
[11-Nov-2010 10:41:18] <Simon4> the user management stuff looked fairly identical from core->enterprise
[11-Nov-2010 10:41:35] <rmatte> yeh, there is "ACL" functionality though that is supposed to limit what each user can see
[11-Nov-2010 10:42:12] <nyeates> yes, there is ACL's.....u can define what devices and reports and alerts that users or groups can have access to
[11-Nov-2010 10:42:20] <nyeates> that is in enterprise
[11-Nov-2010 10:42:22] <rmatte> yeh
[11-Nov-2010 10:42:45] <nyeates> useful for ISPs or people with disparate groups
[11-Nov-2010 10:42:54] <Simon4> ah cool
[11-Nov-2010 10:42:58] * Simon4 goes to see how to do that
[11-Nov-2010 10:43:11] <nyeates> see it in the extended mon guide
[11-Nov-2010 10:46:16] * Simon4 will hassle the trainer when they appear in december
[11-Nov-2010 10:46:43] <rmatte> hehe
[11-Nov-2010 10:46:53] <rmatte> rocket: how goes things on the front line?
[11-Nov-2010 10:47:31] <rmatte> (using front line in a war terminology sense of course)
[11-Nov-2010 10:47:36] <kokey> haha
[11-Nov-2010 10:47:38] <kokey> thanks
[11-Nov-2010 10:47:56] <kokey> i was busy reading through the docs now about users etc. and noticed the good acl stuff is only in enterprise
[11-Nov-2010 10:48:06] <rmatte> yeh
[11-Nov-2010 10:48:15] <kokey> it's a darn pain that enterprise trial is only a windows exe vmware
[11-Nov-2010 10:48:18] <rmatte> which is why I have 15 separate production zenoss instanced hehe
[11-Nov-2010 10:48:22] <rmatte> instances*
[11-Nov-2010 10:48:50] <rmatte> kokey: really?  I'll bet if you bugged them you could get your hands on a different installer.
[11-Nov-2010 10:49:30] <rmatte> ...but if you use it for more than 30 days they'll dispatch an angry gorilla with a big stick to meet up with you
[11-Nov-2010 10:49:43] <nyeates> hah wow....there was a enterprise user who had 12 instances world wide...mostly b/c of geographic lag between the sites.....otherwise, all distributed enterprise is done with remote collectors and hubs
[11-Nov-2010 10:49:45] <kokey> yeah i think what i'll do is let them engage me further and do more vendor song and dance then i'll probably get the installer
[11-Nov-2010 10:50:16] <rmatte> nyeates: yeh, I basically have 1 instance per client for the larger clients, and then some multi-client instances
[11-Nov-2010 10:50:19] <Simon4> nyeates: know if you can have multiple copies of zenactions to get some redundancy in alerting?
[11-Nov-2010 10:50:31] * Simon4 has redundancy issues to solve right now
[11-Nov-2010 10:50:58] <rmatte> Simon4: you mean if the hub goes down?
[11-Nov-2010 10:51:09] <kokey> this is a small site, i think i'll only have about 600 devices in total
[11-Nov-2010 10:51:10] <nyeates> Simon4: multiple instances of zenactions on a single server? is that what u mean?
[11-Nov-2010 10:51:17] <kokey> and could probably do it with about 6 collectors
[11-Nov-2010 10:51:42] <Simon4> nyeates: so two hubs (across disparate geographical locations), zenactions on each hub sending alerts for teh events raised by collectors reporting to those hubs
[11-Nov-2010 10:52:48] <rmatte> Simon4: you'd have the same collectors reporting to both hubs you mean?
[11-Nov-2010 10:52:54] * rmatte doesn't get what you mean
[11-Nov-2010 10:53:33] <Simon4> rmatte: no, just trying to remove the single point of failure that seems to be a single copy of zenactions, and struggling to get my head around how
[11-Nov-2010 10:54:02] <Simon4> everything else can be made redundant somehow, but if we lose the master events/alerts just die
[11-Nov-2010 10:54:02] <rmatte> You could just enable watchdog to watch that daemon
[11-Nov-2010 10:54:08] <rmatte> that way if it fails it'll be automatically restarted
[11-Nov-2010 10:54:11] <Simon4> that does nothing if the network in that DC dies
[11-Nov-2010 10:54:29] <rmatte> true
[11-Nov-2010 10:54:33] <Simon4> daemon failures are the least of my worries
[11-Nov-2010 10:54:39] <nyeates> Simon4: hmmmm, yeah i am not clear on how multiple hubs works out
[11-Nov-2010 10:54:44] <rmatte> why not just have something outside that DC checking up on it?
[11-Nov-2010 10:55:11] <rmatte> you're option for some super complicated solution when the simplest will probably do fine
[11-Nov-2010 10:55:15] <rmatte> opting*
[11-Nov-2010 10:55:27] <Simon4> rmatte: beacause for the time that that network is down, you're completely blind to any events occurring in $other_dc, even though you have all this zenoss infrastructure in there still collecting/monitoring
[11-Nov-2010 10:55:32] <Simon4> which seems totally stupid
[11-Nov-2010 10:55:59] * Simon4 is trying ot make this as simple as possible, whilst still keeping within business requirements, trust me
[11-Nov-2010 10:56:08] <rmatte> How is a second instance of zenactions going to help in that case though?
[11-Nov-2010 10:56:17] <rmatte> are you actually talking about full failover of monitoring?
[11-Nov-2010 10:56:39] <Simon4> rmatte: it may not, I was raising the question to nyeates as he (she?) may  have known a good way to solve this problem
[11-Nov-2010 10:56:44] <Simon4> no, I'm not
[11-Nov-2010 10:56:52] <Simon4> I'm talking about multiple-hub installs that you can do in enterprise
[11-Nov-2010 10:57:04] <Simon4> we already have full failover, it's a PITA
[11-Nov-2010 10:57:11] <rmatte> If the network is down, how do you hope to get the info to the second zenactions daemon?
[11-Nov-2010 10:57:52] <nyeates> Simon4: oh i did not know you were speaking in terms of enterprise....yes this can be done....and now we have a seperate product that will diostribute event creation
[11-Nov-2010 10:57:52] <Simon4> rmatte: that's the question I'm asking
[11-Nov-2010 10:58:19] <Simon4> nyeates: works in 2.5? or 3 only?
[11-Nov-2010 10:58:48] <nyeates> Simon4: I will follow up with you afterwards and get you connected with some ppl who know details....i think it is a 3.0 thing, but ill check
[11-Nov-2010 10:58:56] <Simon4> thanks
[11-Nov-2010 10:59:22] <rmatte> Simon4: carrier pigeons work well as a backup link I hear
[11-Nov-2010 10:59:29] <Simon4>
[11-Nov-2010 10:59:31] <rmatte> muahaha
[11-Nov-2010 10:59:40] <rmatte> they make a terrible mess though
[11-Nov-2010 10:59:45] <nyeates> lol
[11-Nov-2010 11:02:53] <nyeates> Ok everyone
[11-Nov-2010 11:03:30] <kokey> maybe when the hub fails initiate the DR procedure
[11-Nov-2010 11:03:33] <kokey> ;-)
[11-Nov-2010 11:03:33] <nyeates> This will mark my first IRC community developer session with zenoss, without the late Matt Ray
[11-Nov-2010 11:04:06] <nyeates> For those that do not know, I am Nick Yeates at Zenoss Inc. I am taking some of Matt Rays duties
[11-Nov-2010 11:04:14] <nyeates> Matt was our former Community Manager
[11-Nov-2010 11:04:32] <Simon4> kokey: "Wave hands around and scream loudly" ?
[11-Nov-2010 11:04:33] <jrh0090> This will also mark my first IRC developer session as a developer
[11-Nov-2010 11:05:00] <nyeates> yup, thanks for coming Joseph
[11-Nov-2010 11:05:14] <jrh0090> np
[11-Nov-2010 11:05:22] <nyeates> Joseph Hanson has been with Zenoss for the last 8 months and is here to pitch in where he can
[11-Nov-2010 11:06:54] <nyeates> Jospeh, do you know anything about distributed hubs?
[11-Nov-2010 11:07:08] <nyeates> if not, its all good as I am going to point simon at the right resources no matter
[11-Nov-2010 11:07:11] <jrh0090> unfortunately not much, I have worked with distributed collectors a little bit but not really hubs
[11-Nov-2010 11:07:17] <nyeates> k
[11-Nov-2010 11:09:06] <nyeates> so yeah - i didnt see a rush of ppl heading in here for this today. I think it is going to be light. But please, hit us up with your questions, discussions, or favorite lol-cat links
[11-Nov-2010 11:11:05] <rmatte> hehe
[11-Nov-2010 11:13:22] <kokey> I didn't know it was happening, when is the next one?
[11-Nov-2010 11:13:39] <jrh0090> They should be every two weeks
[11-Nov-2010 11:13:41] <nyeates> thurs's every 2 weeks
[11-Nov-2010 11:13:45] <kokey> I'm doing a trial of it here and will probably be oozing questions from about 2 weeks from now
[11-Nov-2010 11:13:50] <kokey> ok, perfect
[11-Nov-2010 11:14:07] <rmatte> well, you can always ooze away when it's not a dev session too
[11-Nov-2010 11:14:25] <kokey> yeah I do anyway, this channel's been incredibly helpful
[11-Nov-2010 11:14:25] <rmatte> aloha kells
[11-Nov-2010 11:14:35] <kells> G'day, g'day!
[11-Nov-2010 11:15:52] <rmatte> kells: wow, I didn't know teksavvy was in the states.  I thought they were a Canadian ISP.
[11-Nov-2010 11:16:29] <kells> They are -- I'm living in Toronto
[11-Nov-2010 11:16:34] <rmatte> ohhh
[11-Nov-2010 11:16:46] <rmatte> didn't realize that
[11-Nov-2010 11:17:01] <rmatte> I'm from Ottawa
[11-Nov-2010 11:17:22] <kells> Oh cool!  I haven't lived there in years, but it's an incredible place in Autumn
[11-Nov-2010 11:17:30] <rmatte> indeed
[11-Nov-2010 11:18:02] <rmatte> So have you been in Toronto a long time?  I always thought you were based out of Austin or something.
[11-Nov-2010 11:18:27] <kells> About 6 months.  Before that I was in Calgary.
[11-Nov-2010 11:18:34] <rmatte> ah right, Calgary
[11-Nov-2010 11:18:53] <rmatte> I knew there was someone in Calgary, forgot it was you
[11-Nov-2010 11:19:47] <kells> Any neat projects on the go?
[11-Nov-2010 11:20:27] <kokey> I was wondering if there is something in the pipeline to make the multi graph reports more idiot friendly
[11-Nov-2010 11:20:30] <rmatte> you mean ZenPack projects?
[11-Nov-2010 11:20:37] <kokey> setting them up that is
[11-Nov-2010 11:21:07] <kells> Well, there's actually some major changes planned for reporting in the next release.
[11-Nov-2010 11:21:43] <jrh0090> I don't think it will affect existing multi-graph reports though, it is more of a replacement
[11-Nov-2010 11:21:47] <kells> The packaging is completely up in the air, but a new capability will be added that will allow you to do very high quality reporting (with multi-tenanting)
[11-Nov-2010 11:22:07] <kells> Right.  The existing reporting in Zenoss will be untouched.
[11-Nov-2010 11:22:53] <kells> It's more of a data warehouse type scheme with the ability to do dashboard type things as well as all sorts of crazy reports.
[11-Nov-2010 11:23:38] <kokey> ah just what i wanted to hear
[11-Nov-2010 11:23:50] <rmatte> kells: I hear none of that may make it in to core?
[11-Nov-2010 11:24:02] <kells> I don't know how they're going to package it.
[11-Nov-2010 11:24:07] * rmatte nods
[11-Nov-2010 11:24:07] <nyeates> it is pretty nice. Packaging and pricing is up in the air
[11-Nov-2010 11:24:14] <kells> I believe that it will be an extra charge option
[11-Nov-2010 11:24:23] <nyeates> likely
[11-Nov-2010 11:24:24] <kokey> and how soon, also up in the air?
[11-Nov-2010 11:24:27] <kells> It *might* be available as a chargeable option for core
[11-Nov-2010 11:24:33] <kells> Keerikes! 
[11-Nov-2010 11:24:52] <kokey> i'm hoping in under 6 months ;-)
[11-Nov-2010 11:24:54] <kells> Avalon is projected for first quarter of 2011
[11-Nov-2010 11:24:58] <jrh0090> I think around the new year
[11-Nov-2010 11:25:04] <kokey> ok perfect
[11-Nov-2010 11:25:13] <jrh0090> yes, first quarter is probably a better answer
[11-Nov-2010 11:25:20] <rmatte> kells: yeh the releases have been pretty close to every 6 months, decent cycle
[11-Nov-2010 11:26:17] <kells> Just finished zenping conversion to the new framework, just in case anybody's interested.
[11-Nov-2010 11:27:00] <kells> Or maybe not.... 
[11-Nov-2010 11:27:31] <kells> It does latency graphing and allows for non-management IP monitoring.
[11-Nov-2010 11:27:48] <kells> Roughly comparable to the existing fping zenpack but more lightweight.
[11-Nov-2010 11:28:13] <Simon4> awesome
[11-Nov-2010 11:28:15] <nyeates> explain non-management IP mon?
[11-Nov-2010 11:28:48] <nyeates> im intersted to understand
[11-Nov-2010 11:28:48] <jrh0090> It also eliminates the need to store routes in the ZODB, right ?
[11-Nov-2010 11:29:37] <kells> As an example, a large-ish router or switch may have a hundred or so network interfaces hanging off it. Right now you pick one IP address to ping, and that's all you get.
[11-Nov-2010 11:30:05] <kells> If you want other interfaces monitored, you need to poll the SNMP status or run an external ping command or rely on SNMP traps
[11-Nov-2010 11:30:59] <kells> As pointed out, the routes are no longer calculated from the model, but are gathered directly from traceroute.  No more "Route xxx not in topology" messages in the zenhub log! (Yay!)
[11-Nov-2010 11:31:03] <nyeates> hi chad
[11-Nov-2010 11:32:28] pmcguire is now known as ptmcg
[11-Nov-2010 11:32:33] <kells> The user interface for MIBs will be slightly spruced up, so you'll be able to add MIBs directly from the GUI.
[11-Nov-2010 11:32:53] <kells> The mib browser functionality probably won't make it into Avalon.
[11-Nov-2010 11:34:23] <kells> zenperfsnmp, zenping, zenstatus and zencommand have all been converted
[11-Nov-2010 11:40:35] <nyeates> nice kells....thats exciting to hear your mib browser might wiggle in eventually
[11-Nov-2010 11:46:07] <davetoo> oh, right   I don't usually get up this early
[11-Nov-2010 11:46:43] <nyeates> hi davetoo
[11-Nov-2010 11:47:06] <davetoo> nyeates: hi; somebody asked about a roadmap the other day.  How's that going?
[11-Nov-2010 11:47:40] <nyeates> that has sparked some internal discussion....it is ongoing and I will be getting back
[11-Nov-2010 11:47:46] <davetoo> *nod*
[11-Nov-2010 11:48:37] <kells> Is there anything in particular you're interested in?
[11-Nov-2010 11:49:40] <davetoo> I think the question was prompted by a blog post I made about RelStorage; I do know the story with that, through off-channel discussions.
[11-Nov-2010 11:49:56] <davetoo> The next thing I'm interested in is the new ZEP code in svn
[11-Nov-2010 11:50:09] <kells> Uh oh
[11-Nov-2010 11:50:14] <davetoo> I think that is there mostly to support an Enterprise feature
[11-Nov-2010 11:50:16] <davetoo> heh
[11-Nov-2010 11:50:47] <jrh0090> Zep is the new event processing system. It is a huge change with the goal of processing several times more events then we currently do
[11-Nov-2010 11:50:57] <ptmcg> ZEP will be part of core, offloading compute-intensive CPU work from zenhub
[11-Nov-2010 11:51:26] <davetoo> transforms and such?
[11-Nov-2010 11:51:49] <davetoo> Will it help zensyslog performance?
[11-Nov-2010 11:51:50] <jrh0090> transforms will be taken care of by a new daemon
[11-Nov-2010 11:52:37] <davetoo> i.e. will any of the work that zensyslog does be moved/offloaded to zep?
[11-Nov-2010 11:52:39] <kells> What's the issue with zensyslog performance? Losing messages internally, or sending messages to zenhub?
[11-Nov-2010 11:52:56] <davetoo> in the past it's been a huge CPU hog for me.
[11-Nov-2010 11:53:11] <ptmcg> wow, people poke around in svn?
[11-Nov-2010 11:53:14] <kells> With zensyslog, there are some new daemon options (in 3.0) that allow you to get more buffer space to deal with losing messages.
[11-Nov-2010 11:53:35] <ptmcg> I'll have to be careful about my checkin comments...
[11-Nov-2010 11:53:37] <davetoo> ptmcg: I don't know how many people do, but yeah, I do
[11-Nov-2010 11:53:39] <kells> zensyslog also allows you to drop messages based on regexes, so that might be an approach too.
[11-Nov-2010 11:54:19] <davetoo> regexes can be slow if not carefully written, of course.  Sometimes substring ops are faster.
[11-Nov-2010 11:54:38] <davetoo> anyway, good to hear about zep.  Thanks.
[11-Nov-2010 11:55:08] <kells> In your case, were you able to isolate the cause of the CPU load?
[11-Nov-2010 11:55:10] <davetoo> ptmcg: in fact I'm trying to clone it to Git,
[11-Nov-2010 11:55:36] <davetoo> kells: no, and project needs dictated that I defer that problem; I'm no longer working at that firm.
[11-Nov-2010 11:55:49] <kells> Ah, okay
[11-Nov-2010 11:57:38] <kells> I hadn't seen any issues with CPU performance and zensyslog. Usually it's an issue with losing messages due to buffer space issues.
[11-Nov-2010 11:57:51] <kells> And regular syslogd has issues with that too
[11-Nov-2010 11:57:58] <rmatte> the simplest way to deal with high load on zensyslog is to put syslog-ng before it, have it filter out most of the garbage and forward the rest to Zenoss
[11-Nov-2010 11:58:11] <davetoo> rmatte: which is what I was going to do, yes.
[11-Nov-2010 11:59:52] <davetoo> Can you talk about how AMQP/Rabbit, protobuffers, and zep fit together?
[11-Nov-2010 12:00:20] <davetoo>
[11-Nov-2010 12:00:25] <ptmcg> !!!
[11-Nov-2010 12:00:25] <jrh0090> That is a loaded question
[11-Nov-2010 12:00:30] <davetoo> hehehehe
[11-Nov-2010 12:00:57] <jrh0090> In a nutshell we are using rabbitmq to communicate from zenhub to ZEP
[11-Nov-2010 12:01:15] <davetoo> I've been looking at the code from the standpoint of HA and distributed processing
[11-Nov-2010 12:01:16] <jrh0090> events will hang out in the queue until they are processed by zep
[11-Nov-2010 12:01:22] <davetoo> ah
[11-Nov-2010 12:01:26] <jrh0090> this allows us to handle event storms and high loads better
[11-Nov-2010 12:01:50] <jrh0090> the protobufs are an IDL (interface definition language) so we can send objects accross the wire instead of strings
[11-Nov-2010 12:01:52] <kokey> ok, i have very important drinking to do
[11-Nov-2010 12:01:52] <davetoo> aye
[11-Nov-2010 12:01:55] <kokey> bye everyone
[11-Nov-2010 12:01:57] <Simon4> seu
[11-Nov-2010 12:02:21] <ptmcg> kokey: is there unimportant drinking?
[11-Nov-2010 12:02:34] <davetoo> interesting
[11-Nov-2010 12:02:35] <kokey> yeah sometimes it's just because i'm thirsty
[11-Nov-2010 12:02:38] <rmatte> not when I'm around
[11-Nov-2010 12:02:40] <rmatte> lol
[11-Nov-2010 12:03:11] <tiredofme> has anyone had success with application dependencies? essentially, don't alert me about my websites if the db backend dies for the webservers
[11-Nov-2010 12:03:54] <rmatte> tiredofme: didn't you implement that transform and it worked?
[11-Nov-2010 12:04:41] <tiredofme> rmatte: yep, it seemed to work for the local device that it applies to. I'm just wondering if anyone had examples of making that sort of transform work across multiple systems
[11-Nov-2010 12:05:03] <rmatte> I designed the transform to be general
[11-Nov-2010 12:05:07] <rmatte> it should apply to anything
[11-Nov-2010 12:05:34] <rmatte> but if it's not working, you need to step back and examine the logic
[11-Nov-2010 12:05:39] <rmatte> forget about python, pseudo code it out
[11-Nov-2010 12:05:45] <rmatte> use a flow chart if you have to
[11-Nov-2010 12:06:02] <rmatte> then once you have the logic figured out, examine the transform and make the necessary modifications
[11-Nov-2010 12:06:41] <tiredofme> got it
[11-Nov-2010 12:07:19] <ptmcg> what is the nastiest, ugliest transform you've ever written?
[11-Nov-2010 12:07:48] <rmatte> ptmcg: in terms of complexity?
[11-Nov-2010 12:08:06] <rmatte> I never write "ugly" transforms, they are works of art
[11-Nov-2010 12:08:17] <rmatte>
[11-Nov-2010 12:08:17] <ptmcg> complexity, external dependencies, whatever
[11-Nov-2010 12:08:24] <davetoo> I wish there were an option to upload transforms as a file; I'd rather keep them versioned
[11-Nov-2010 12:08:28] <rmatte> ah, I stay away from external dependencies
[11-Nov-2010 12:08:46] <rmatte> actually there was one...
[11-Nov-2010 12:09:05] <Simon4> davetoo: someone I know might have written a zenpack that wraps up transforms in svn and delivers them as zenpack contents
[11-Nov-2010 12:09:12] <Simon4> so you can do just that
[11-Nov-2010 12:09:19] <davetoo> aye
[11-Nov-2010 12:09:48] <rmatte> trying to dig it up...
[11-Nov-2010 12:09:59] <ptmcg> ah... a good story is coming...
[11-Nov-2010 12:10:11] <davetoo> ?
[11-Nov-2010 12:10:35] <rmatte> http://dev.zenoss.org/trac/ticket/5376
[11-Nov-2010 12:10:40] <rmatte> the transform in that trac ticket
[11-Nov-2010 12:10:57] <rmatte> the goal was to actually have the operational status represent the status of interfaces pre 2.5
[11-Nov-2010 12:11:07] <rmatte> before /Status/IpInterface was introduced
[11-Nov-2010 12:11:13] <rmatte> (which still isn't created by default)
[11-Nov-2010 12:11:41] <rmatte> but it didn't work since Zenoss would stop monitoring the interface after the status was changed
[11-Nov-2010 12:12:31] <rmatte> I updated this to reflect the use of /Status/IpInterface: docs/DOC-2494
[11-Nov-2010 12:13:21] <rmatte> also, all that import stuff was necessary at the time
[11-Nov-2010 12:13:31] <rmatte> it isn't anymore since dmd is now available in transforms by default
[11-Nov-2010 12:13:43] <ptmcg> The device is too
[11-Nov-2010 12:13:54] <rmatte> oh?
[11-Nov-2010 12:13:58] <rmatte> I wasn't aware of that
[11-Nov-2010 12:14:03] <rmatte> is it just "d"?
[11-Nov-2010 12:14:05] <Simon4> dev and evt
[11-Nov-2010 12:14:06] <rmatte> or is it dev?
[11-Nov-2010 12:14:10] <ptmcg> The lookup from evt.ipAddress -> device should already be done, I believe
[11-Nov-2010 12:14:12] <rmatte> ah ok
[11-Nov-2010 12:14:14] <rmatte> good to know
[11-Nov-2010 12:14:16] * Simon4 didn't know dmd was avail - is that a new thing for 3?
[11-Nov-2010 12:14:19] <ptmcg> dev and device are both defined
[11-Nov-2010 12:14:39] <rmatte> Simon4: nah, new thing for 2.5
[11-Nov-2010 12:14:40] <ptmcg> I don't think it is anything new
[11-Nov-2010 12:14:44] <Simon4> rmatte: ah cool
[11-Nov-2010 12:14:47] <rmatte> it wasn't available in 2.
[11-Nov-2010 12:14:49] <rmatte> 2.4*
[11-Nov-2010 12:16:44] <davetoo> back on the dependency thing: how/when is the pingtree used for layer 3 dependencies...
[11-Nov-2010 12:17:06] <davetoo> I see some bits in the code about polling for device ping issues
[11-Nov-2010 12:17:15] <nyeates> ive been here since 2.3.3  at least, and i *thought* dmd has been around
[11-Nov-2010 12:17:46] <rmatte> nyeates: I started out with 2.3.2 as my first version
[11-Nov-2010 12:17:55] <ptmcg> 'log' is also defined
[11-Nov-2010 12:18:02] <rmatte> I wasn't able to use dmd in a transform without importing it before 2.5
[11-Nov-2010 12:18:13] <rmatte> actually, there was a patch to allow that, let me find it...
[11-Nov-2010 12:18:21] <nyeates> ohh, i misunderstood
[11-Nov-2010 12:18:42] * nyeates slaps himself witha  wet fish
[11-Nov-2010 12:19:11] <rmatte> Patch 14049 - Adds support for the use of "dmd" in transforms (as you would use it in zendmd).
[11-Nov-2010 12:19:24] <davetoo> but I'm not clear on whether all daemons stop testing/polling based on that, and... I'm not sure if there's an Acquisition path or something that goes up the pingtree to see if an upstream router is down
[11-Nov-2010 12:20:59] <davetoo> i.e. does getDevicePingIssues(?) see only that device, or is that polling a property that's part of an Acquisition tree?
[11-Nov-2010 12:21:26] <rmatte> davetoo: in my experience the layer 3 dependencies are extremely flaky.  I've never gotten them working.
[11-Nov-2010 12:21:31] <davetoo> (I'm still not sure I understand that Acquisition black magic)
[11-Nov-2010 12:22:06] <davetoo> rmatte: that's why I'm asking these questions   It seems like there should be some observer patterns to invalidate subtrees as soon as a router goes down...
[11-Nov-2010 12:22:22] <rmatte> yup
[11-Nov-2010 12:23:55] <rmatte> I still vote for manual device dependencies to be implemented
[11-Nov-2010 12:24:01] <rmatte> I'd rather just define what is what myself
[11-Nov-2010 12:24:21] <rmatte> automation is fine, but not always effective
[11-Nov-2010 12:30:10] <mloven> any zendmd ninjas out there today? 
[11-Nov-2010 12:30:43] <rmatte> what do you need?
[11-Nov-2010 12:31:07] <mloven> I'm looking for a zendmd script that will display the number of devices of a particular type, per device group...
[11-Nov-2010 12:31:14] <mloven> so for example:
[11-Nov-2010 12:32:42] <mloven> Group1 has a router at /Devices/Network/Router and 2 routers at Devices/Network/Cisco/Router.  The zendmd script should display something like:
[11-Nov-2010 12:33:01] <mloven> Group Name    Routers
[11-Nov-2010 12:33:02] <mloven> Group1            3
[11-Nov-2010 12:33:13] <mloven> crazy right? 
[11-Nov-2010 12:33:26] <rmatte> that's not difficult to do BUT...
[11-Nov-2010 12:33:43] <rmatte> /Devices/Network/Cisco/Router is just an organizer
[11-Nov-2010 12:33:57] <rmatte> It could just as easily be /Devices/Network/Cisco/ClownPants
[11-Nov-2010 12:33:59] <mloven> yep.
[11-Nov-2010 12:34:10] <Simon4> is "Group1" an actual group in zenoss?
[11-Nov-2010 12:34:24] <mloven> for the purposes of this example, yes...
[11-Nov-2010 12:34:26] <rmatte> do you have a system where the last organizer in the device classes is always the device type?
[11-Nov-2010 12:35:19] <rmatte> Honestly, it would be easier for you to use Systems and define Router, Switch, Firewall, etc... in there
[11-Nov-2010 12:35:27] <rmatte> then assign the devices to whatever system group
[11-Nov-2010 12:35:28] <mloven> rmatte: no, in some cases, we have, say /Devices/Network/Cisco/Router/Voice Gateway, or similar
[11-Nov-2010 12:36:05] <rmatte> because your device class structure isn't going to be as concrete/accurate
[11-Nov-2010 12:36:44] <mloven> rmatte:  that's a doable option
[11-Nov-2010 12:36:52] <rmatte> then that would be your best option
[11-Nov-2010 12:37:06] <rmatte> much easier to just query systems groups than to try and do a bunch of magic
[11-Nov-2010 12:37:07] <davetoo> however,
[11-Nov-2010 12:37:20] <davetoo> iit at used to be the case that /Devices/Network/Router was special
[11-Nov-2010 12:37:26] <davetoo> is it no longer the case?
[11-Nov-2010 12:37:33] <rmatte> "Special"?
[11-Nov-2010 12:37:36] <rmatte> in what sense?
[11-Nov-2010 12:37:50] <rmatte> It's just a class
[11-Nov-2010 12:37:59] <davetoo> Routers were treated differently...
[11-Nov-2010 12:38:10] <davetoo> maybe that was only a zendisc thing
[11-Nov-2010 12:38:11] <rmatte> maybe based on whatever plugins are defined in that class
[11-Nov-2010 12:38:15] <mloven> like "short-bus" special? 
[11-Nov-2010 12:38:23] <davetoo> no, not in the traditional templates sense
[11-Nov-2010 12:38:25] <rmatte> nothing to keep you from creating your own class and defining the same zProperties/Plugins
[11-Nov-2010 12:38:38] <davetoo> rmatte: right, I'm not talking about that stuff
[11-Nov-2010 12:38:51] <rmatte> well I don't see what else would apply
[11-Nov-2010 12:38:56] <davetoo> but.. my CRS is preventing me from remembering if it was anything more than how zendisc treated them.
[11-Nov-2010 12:38:58] * rmatte listens intently
[11-Nov-2010 12:39:13] <davetoo> I'll find my question in the mail archives later
[11-Nov-2010 12:39:17] <davetoo> busy righ tnow
[11-Nov-2010 12:39:17] <rmatte> k
[11-Nov-2010 12:39:26] <davetoo> you could search
[11-Nov-2010 12:39:36] <davetoo> it's old enough that either EAD or ECN answered
[11-Nov-2010 12:40:36] <mloven> either way, the "system" approach should circumvent any issues, right?
[11-Nov-2010 12:41:34] <rmatte> yessir
[11-Nov-2010 12:44:14] <mloven> ok.  done deal then.   given that option, how would one go about squeezing zendmd to give those results?
[11-Nov-2010 12:46:26] <rmatte> the number of devices in each?
[11-Nov-2010 12:46:42] <mloven> yeah, just a count.
[11-Nov-2010 12:46:46] <rmatte> pretty simply...
[11-Nov-2010 12:46:52] <mloven> but per group
[11-Nov-2010 12:47:50] <rmatte> yeh, give me a minute
[11-Nov-2010 12:48:45] <nyeates> BTW, we will announce Summer of Zenpack winners Mon morn, likely
[11-Nov-2010 12:48:53] <rmatte> coolio
[11-Nov-2010 12:49:07] <nyeates> there really are some cool ones - and i hope that those not scouring our ZPs can use the winners list as a "best of the best" list
[11-Nov-2010 12:49:10] <rmatte> If I win I'm picking the porsche 911 as my prize
[11-Nov-2010 12:49:24] <nyeates> weve got 16 prizes we are giving away, and a few of you in here are winners :-)
[11-Nov-2010 12:49:37] <nyeates> we ran out of those, sorry
[11-Nov-2010 12:49:40] * Simon4 wants the aston martin
[11-Nov-2010 12:49:50] <rmatte> I had 6 submissions, 3 of which were new original packs, so I hope that I've at least won something
[11-Nov-2010 12:50:30] <nyeates> egor and andrea had most submissions i think....probably you were 3rd ryan
[11-Nov-2010 12:50:40] <rmatte>
[11-Nov-2010 12:50:43] <rmatte> Egor is a machine
[11-Nov-2010 12:51:12] <rmatte> which reminds me, I have an enhancement planned for his WMI Performance pack
[11-Nov-2010 12:51:19] <rmatte> just need to find time to do it
[11-Nov-2010 12:51:37] <nyeates> cool
[11-Nov-2010 12:52:06] <rmatte> It's going to allow it to monitor windows mount points
[11-Nov-2010 12:52:14] <rmatte> instead of just logical filesystems
[11-Nov-2010 12:52:46] <nyeates> i think that that is oft requested feature when i was in support
[11-Nov-2010 12:53:16] <rmatte> mloven: I'm almost done with this dmd script, don't worry, haven't forgotten lol
[11-Nov-2010 12:53:41] <rmatte> mloven: so you just want a count of each type, but per group?
[11-Nov-2010 12:53:51] <rmatte> So something like:
[11-Nov-2010 12:53:54] <rmatte> Group 1:
[11-Nov-2010 12:53:57] <rmatte> Network: 4
[11-Nov-2010 12:53:59] <rmatte> Security: 0
[11-Nov-2010 12:54:02] <rmatte> or whatever
[11-Nov-2010 12:54:03] <rmatte> ?
[11-Nov-2010 12:54:42] <rmatte> nyeates: yeh, it's being requested by a large client of ours (who apparently has their own instance of Zenoss running internally)
[11-Nov-2010 12:54:57] <rmatte> nyeates: but they lack a Zenoss wizard to get that level of monitoring in place
[11-Nov-2010 12:55:10] <rmatte> Looking at the WMI queries it's going to be very simple to implement
[11-Nov-2010 12:55:23] <rmatte> just going to use Win32_Volume instead of Win32_LogicalDisk
[11-Nov-2010 12:55:28] <rmatte> since it has all the same values
[11-Nov-2010 12:55:34] <rmatte> but also provides values for the mount points
[11-Nov-2010 12:55:49] <mloven> rmatte:  yeah, so how many routers Group1 has, how many routers Group2 has, and so on.
[11-Nov-2010 12:55:54] <rmatte> and I checked where egor is already pulling the filesystem IO info from, it supplies info for the mount points too
[11-Nov-2010 12:56:10] <rmatte> mloven: k, 1 more min and I'll have it for you
[11-Nov-2010 12:56:24] <ptmcg> I've seen some problems with collecting parameters from Win32_Volume - if you try this please send in your results, either success or the other kind
[11-Nov-2010 12:57:15] <rmatte> ptmcg: will do
[11-Nov-2010 12:57:29] <rmatte> I examined the values and they all checked out, so hopefully it'll work as expected
[11-Nov-2010 13:00:10] <rmatte> mloven: hmmm, it'll take a few more minutes, just hit a small sticking point
[11-Nov-2010 13:04:46] <mloven> rmatte:  np, I appreciate the help
[11-Nov-2010 13:05:13] <davetoo> Thanks, dev guys
[11-Nov-2010 13:05:33] <nyeates> yes thx to all!
[11-Nov-2010 13:06:54] <nyeates> Im gonna duck out now and change the title....later everyone
[11-Nov-2010 13:07:44] <rmatte> later
Comments (0)