monit-general
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: m/monit - centralized monitoring


From: Mark Ma
Subject: Re: m/monit - centralized monitoring
Date: Fri, 7 Nov 2003 06:20:21 +0800

>Jan-Henrik Haukeland<address@hidden> writes:
>> "Mark Ma" <address@hidden> writes:
> 
> > This is Mark from Shanghai,China. first of all , i am very excited
> > to know that u r in the process of such a monitor system.
> >
> > actually, i am in this process to ,wish to have more comunication
> > with u to talk about that. maybe i could get some good suggestiong
> > from you.
> 
> Heh, seems that there are many with this idea nowdays :-) I recently
> discussed this with Georges Toth also (who is a subcriber to this list
> and a monit user). In the monit project we have also discussed this
> idea now and then in the past.
> 
> #
> 
> I sincerely belive that building a centralized monitor application
> upon monit as an "agent" is an interesting twist; The thing is, that
> most, if not all of the existing centralized monitoring systems out
> there, such as HP OpenView, Nagios and Big Brother, just to mention a
> few, starts in the "other end". That is; they start with the central
> application and add "thin" agents or plugins on an ad-hoc basis (or so
> it seems). The result is often a centralized application with a nice
> GUI and with lots of statistical data and with lots if fine graphs,
> but quite often with an underlying "rotten" architecture. Configuring
> and setting up these systems are usually quite complicated, not at
> least because of the myriad of plugins.
##

i really appreciate some 'BIG & Terrible' system ,such as HP VPO.coz u know,i 
have been delivery some VPO system for my customer.In my opinion ,the 
infrastructure of VPO is really pretty good. Most actions of its agent is using 
shell scripts and some in C code.i think,if there r many monited objet , C code 
could lowdown avarage load of system .and,  in my opinon ,if u wanna know more 
detail info about your OS || application,you must use hard code programing . 
'centralized monitor system' has difference with 'Monitor & Prediction system' 
.the more information u could get ,the more prediction u could make ,am i right 
? well, if one administrator have no enough technical ability, even script will 
make him headache  ,i think. so , the first , administrator should have enough 
technical experience , the second ,i still appreciate a program-based monitor 
system,even though it's hard for different system delivery(program porting 
should be a big headache problem ,lol).

in my designer , i will use scripts as less as possible .(So ,u maybe think ,i 
will not in GPL permission later. :-| ) 

while ,in the other hand ,as u say ,monitor system ,like VPO ,it 's really hard 
& complicated for admin. for example , communication between the management 
server & managed node with rpc (in some configuration ,VPO could think the node 
is DIED if rpc communication failed ,GOSH ...),so many useless template,bad 
openview database infrastructure ,and some un-optimized SQL phrase,and so on.

then ,i think one monitor system should have a small core ,and easyly extended 
function, maybe like BB and monit.but i really dont know why BB and monit use 
many shell script ? just for easy realizing the function ? if u have a HP 
Superdome ,maybe u dont worry about that . :) 

try to think of this question ,if you wanna monitor 20 
objects(CPU/MEM/DISK/PS/PV/LV and so on ) in one of your target node , maybe u 
could use script to collect the statics what u wanna . BUT, how about  200 
object in your target node? what do u think about perfomance then ?  i could 
imagine that situation---MASS OF SCRIPT s' WORLD ,i think.

> Staring at the other end with monit is interesting, because monit is
> not a "thin" agent but a powerful application in it's own right and
> can do lots of useful things on localhost. This way, a centralized
> monitoring application need not to be very complicated, instead of
> having many thin plugins we can instead have one thin centralized
> application which simply is an aggregated view against many monit
> instances running on many hosts. Besides if you run monit on more than
> one server it will be quite useful to have one and only one
> web-application from where you can control the many monit instances.
> 
> Oh well, that is the sales pitch :)
>

u r a good MONIT sales  . :)

> > in the other hand, i am not sure if your system is being developed
> > by yourself or your company ? if it's yourself's session ,so could i
> > join in with you ?
> 
> Hmm, I originally planned on doing this for my company but with a very
> flexible license, i.e. free for non-commercial use and so on
> (something like the Big Brother license) but I'm very ambivalent on
> this. I haven't really decided yet, but most likely (90%) I will go
> for a GPL license but retain the copyright for my company (also for
> any third-party contributions so I at least have the option to do a
> dual license).

maybe i gonna develop semi opened system ,which the core will be in 
licensed.and the plug-in will be free.the plug-in have binary and scripts 
.while most of mine is binary program.
 
> Still the system will not be completely open-source because I plan to
> build it as C servlets on a C application server called zild
> (http://zild.no/). This application server is closed source and built
> by my company but has a license so that GPL code can use zild. (It's
> similar to building a Java application and release the application
> under GPL altough the Java platform itself is Sun properitary.)

> Anyway, first I have to build and release a first version of the
> centralized monit system and if you like the architecture and code you
> are very welcome to contribute ideas and code (this invitation is of
> course open to anyone and everyone).
> 
> If you or others instead will want to start your own project and build
> your own centralized monit application based on monit I have nothing
> against that of course :) (although forking is seldom a good thing).
> 
sometimes forking  will apprea a new world . :)  

i have been using your monitor system in your website .really good! while i 
find it's just a 'MONITOR' system.what do u think of the history data's analyze 
& predictor of feture grow ?for HP-UNIX ,if i wanna know if i-node is full(it's 
also dangerous) , what info can i get from your monitor UI ?

i think ,monit is focus on application status more than other critical detail . 
if i was an administrator ,my customer will kick me before the monitor system 
tell me 'hi ,ur Oracle database has been crashed!', maybe. :) so ,i think we 
should think about what  an administrator really need ,and what a system should 
be checked and monitored .do u think it's simply some application ? what's your 
monitor system's AIM on hell ? 

> > P.S. i cant have a connection with your web site (http://zild.no)
> > ,is it working now ?
> 
> Yes, it should be up. (The server behind http://zild.no is of course
> zild and it's not impossible that the server was down a minute or two
> this night but monit is of course running on the server so it goes up
> again :)
> 
> -- 
> Jan-Henrik Haukeland
> 


that's all in my opinon ,dont mind what i say ,i just wanna talk about what we 
should do and what's the way we should do ? maybe i have something wrong . :)

thanks.

Mark Ma



> --
> To unsubscribe:
> http://mail.nongnu.org/mailman/listinfo/monit-general
> 

reply via email to

[Prev in Thread] Current Thread [Next in Thread]