On Thursday 16 May 2002 10:07 pm, Brian Fahrlander wrote:
> On Thu, 16 May 2002 21:14:34 -0400, "Skip Gaede" <[EMAIL PROTECTED]> wrote:
> > Brian,
> >
> > You clearly know your SNMP stuff!
>
> Oh, no...I'm just a wanderer bangin' on things until they work. I've
> been exposed to SNMP several times and it just never really sank in.
>
> > In going through the EXAMPLE.conf file
> > supplied with 4.2.5 ucd-snmp from sourceforge, it looks like we might be
> > able to collect the raw data needed from the clients by capturing the
> > data in three tables maintained in the /proc filesystem
> >
> > /proc/stat
> > /proc/meminfo
> > /proc/net/dev
> >
> > I added the following three lines to your config file:
> >
> > exec .1.3.6.1.4.1.2021.50 meminfo /bin/sh /tmp/meminfo.sh
> > exec .1.3.6.1.4.1.2021.51 kernelstats /bin/sh /tmp/kernelstats.sh
> > exec .1.3.6.1.4.1.2021.52 netinfo /bin/sh /tmp/netinfo.sh
> >
> > Each of the scripts "cats" the corresponding file in the /proc
> > filesystem, and the data can be obtained with the command
> >
> > snmpwalk -v 1 ws004 public .1.3.6.1.4.1.2021.50
> >
> > [skip@sgaede skip]$ snmpwalk -v 1 localhost private .1.3.6.1.4.1.2021.50
> > enterprises.2021.50.1.1 = 1
> > enterprises.2021.50.2.1 = "meminfo"
> > enterprises.2021.50.3.1 = "/bin/sh /tmp/meminfo.sh"
> > enterprises.2021.50.100.1 = 0
> > enterprises.2021.50.101.1 = " total: used: free: shared: \
> > buffers: cached:"
> > enterprises.2021.50.101.2 = "Mem: 261877760 257609728 4268032 0
> > \ 16596992 67489792"
> > enterprises.2021.50.101.3 = "Swap: 254943232 246378496 8564736"
> > enterprises.2021.50.101.4 = "MemTotal: 255740 kB"
> > enterprises.2021.50.101.5 = "MemFree: 4168 kB"
> > enterprises.2021.50.101.6 = "MemShared: 0 kB"
> > enterprises.2021.50.101.7 = "Buffers: 16208 kB"
> > enterprises.2021.50.101.8 = "Cached: 61908 kB"
> > enterprises.2021.50.101.9 = "SwapCached: 4000 kB"
> > enterprises.2021.50.101.10 = "Active: 45356 kB"
> > enterprises.2021.50.101.11 = "Inactive: 184456 kB"
> > enterprises.2021.50.101.12 = "HighTotal: 0 kB"
> > enterprises.2021.50.101.13 = "HighFree: 0 kB"
> > enterprises.2021.50.101.14 = "LowTotal: 255740 kB"
> > enterprises.2021.50.101.15 = "LowFree: 4168 kB"
> > enterprises.2021.50.101.16 = "SwapTotal: 248968 kB"
> > enterprises.2021.50.101.17 = "SwapFree: 8364 kB"
> > enterprises.2021.50.101.18 = "NrSwapPages: 2091 pages"
> > enterprises.2021.50.102.1 = 0
> >
> >
> > I did the work on the server, and am now compiling the code for the
> > clients. What do you think: is this sufficient or should we create a
> > custom MIB?
>
> Actually, there are two issues here:
>
> 1. All this information is a part of the MIB that comes with Redhat.
> There's even a means to enumerate the information- one entry is a process
> number, and there's a corresponding entry for that process's shared memory,
> system memory and all sorts of things I didn't think I'd ever use,
> anywhere...but here. I'm a big fan of relying on previous work where it
> exists, and where it allows us the answers we need. It's all there...let's
> not burden the system being monitored with any more computation than it
> needs.
>
> 2. SNMP by nature, is very low level. They don't send ASCII strings
> unless it's the only way. You generally get numbers, not pretty-printed
> names identifying them, and making the job of a remote computer making
> sense of it all that much easier. and if/when the results of cat'ing these
> structures changes, we won't have to publish duplicate-looking releases, ya
> know?
>
> I'd say (for systems using Redhat and/or Mandrake) we stick to the
> original standard...the last thing we need is YET ANOTHER confusing
> standard to adhere to, and to wonder which is in use.
>
> Take another look at the results of the snmpwalk generated from my
> configfile. There are **boatloads** of information there, all laid out so
> that perl programmers and such would be able to loop through'em and get all
> the details in as few transfers as possible. Look close though- there's a
> LOT of data there, when you walk it all.
>
Brian (& Jim),
I'm OK with getting the data any way it works. The point you raise about the
amount of available data is correct: there's so much in the sandbox I have no
clue as to where to start looking. In the case of the stuff in the /proc
"files", I do know what it means, or, where to go to get the description. I
need help getting from what I currently know to getting the same data using
the existing MIBs. (I think the performance stuff is in the UCD-SNMP-MIB, and
the networking stuff is in the IF-MIB. How would one grab that data?)
BTW, yesterday was my first attempt at doing anything with SNMP, and I now
can poke at an SNMP server on a Mac client from my AMD server. Is it worth
defining a custom MIB, or should I just go with bits and pieces from the
already defined stuff?
Thanks,
Skip
_______________________________________________________________
Have big pipes? SourceForge.net is looking for download mirrors. We supply
the hardware. You get the recognition. Email Us: [EMAIL PROTECTED]
_____________________________________________________________________
Ltsp-discuss mailing list. To un-subscribe, or change prefs, goto:
https://lists.sourceforge.net/lists/listinfo/ltsp-discuss
For additional LTSP help, try #ltsp channel on irc.openprojects.net
