[hobbit] Hobbit newbie from BB: differences and what may I lose from migrating?
Camp, Neil D. (ManTech) CTR
neil.camp at deca.mil
Wed Aug 2 13:37:30 CEST 2006
This is an awesome write up. Maybe you should consider adding it to your
webpage?
-----Original Message-----
From: Henrik Stoerner [mailto:henrik at hswn.dk]
Sent: Tuesday, August 01, 2006 5:13 PM
To: hobbit at hswn.dk
Subject: Re: [hobbit] Hobbit newbie from BB: differences and what may I
lose from migrating?
Hi Jordan,
I'll try to answer your questions. Since I also develop Hobbit I am
probably slightly biased when it comes to the "is-this-more-difficult-
to-do-than-with-BB" type of questions, but I am sure others will
voice their opinions on that.
On Tue, Aug 01, 2006 at 12:36:29PM -0700, Jordan Mendler wrote:
>
> First, after reading through whatever I could find on the website I am
> still a little bit confused about configuration and setup. With BB,
you
> install and configure each client and server on the local machine,
> except for the universal bb-hosts. Is this the same on Hobbit, or does
> Hobbit use a central configuration file that is modified only on the
> server to configure clients? I am trying to figure out the difference
> between installing, maintaining and configuring BB and Hobbit setups.
First, let me stress that Hobbit is fully compatible with your existing
BB clients. You can keep your current client setup and just switch to
Hobbit on the server side, and all of your clients will continue to
work as they do with BB as the server. So you can migrate the server
side first, and then migrate clients when you find that it is convenient
to do so - or you want to take advantage of some of the new stuff that
is in Hobbit.
The Hobbit client configuration is maintained on the Hobbit server.
Clients in Hobbit are designed to be *really* dumb; they just collect
data, and all of the configuration of what to monitor, what thresholds
to use for e.g. disk utilisation and so on is configured only on the
Hobbit server.
This is a major difference between Hobbit and BB. With BB you have
delegated the client administration to whoever manages each server.
Hobbit centralizes the monitoring configuration, so you will probably
have a group of people who take more control of the monitoring setup.
> Hobbit looks alot more complex to setup, but once I get my feet wet is
> it any harder than BB?
I think it is easier, once you get used to the Hobbit way of doing
things. But as I said, I am biased.
> Second is performance. I know this list may be biased toward Hobbit,
but
> is it actually faster? We have about 50-100 clients on BB and I did
not
> notice any performance issues.
With that number of systems monitored, you probably will not see a huge
difference. BB works quite well for a small number of systems, but when
you move beyond a couple of hundred boxes the overhead of generating
webpages through shell scripts becomes very noticeable. On my setup,
the servers were simply choking on the disk I/O caused by BB saving
every status in a separate file, and from the huge number of small
cut-grep-awk-sed etc. commands that ran to generate webpages.
> Hobbit looks like it is very complex, so does this mean it uses a lot
> of resources on the client and server? What speed/ram server is
usually
> the minimum recommended for a dedicated Hobbit server? Would something
> like a dual Pentium II 266mhz have any performance issues as a server,
> if it does nothing else? What about for clients? We have still have
> some testing, stating and production servers left that are singe chip
> Pentium III 700-850 mhz, and even a couple Pentium II's. Just need to
> make sure all the resources used for things like graphs are taken from
> the server and not each client.
The Hobbit server uses fewer ressources than the BB server. The main
ressource usage is memory; Hobbit keeps everything in memory except
the history logs and the RRD files used for graphs. That doesn't mean
a whole lot, though: Here's a ps listing of the Hobbit processes running
on my main monitoring system - it handles about 2500 hosts:
$ ps vax|cut -c1-100|egrep "PID|hobbit"
PID TTY STAT TIME MAJFL TRS DRS RSS %MEM COMMAND
732 ? Ss 1:24 0 101 1802 696 0.0 hobbitlaunch
735 ? S 2434:37 1 162 31357 29784 2.8 hobbitd
1470 ? S 14:50 0 99 2332 1088 0.1 hobbitd_channel
--channel=stachg
1471 ? S 25:18 0 108 2515 1048 0.1 hobbitd_history
1472 ? S 964:26 0 99 2332 1264 0.1 hobbitd_channel
--channel=page
1473 ? S 1227:34 0 154 5661 3912 0.3 hobbitd_alert
1474 ? S 4090:05 0 99 2332 1264 0.1 hobbitd_channel
--channel=status
1475 ? D 2962:15 0 178 7381 4392 0.4 hobbitd_rrd
1476 ? S 259:55 0 99 2332 1208 0.1 hobbitd_channel
--channel=data
1477 ? S 494:13 0 178 5141 2128 0.2 hobbitd_rrd
1478 ? S 126:20 0 99 2844 1832 0.1 hobbitd_channel
--channel=client
1480 ? S 291:20 0 146 4485 2792 0.2 hobbitd_client
5552 ? S 0:00 0 669 2002 1352 0.1 sh -c vmstat 300
2 1>/usr/lib/hobbit/client/
As you can see, the biggest chunk of memory goes to the "hobbitd"
process which is the one that keeps all state information. It's
currently using some 31 MB of memory. (This box has 1 GB RAM).
A rough estimate of how much memory Hobbit needs would be the size of
your bbvar/logs/ directory, plus 30 MB.
As for CPU usage, your PII/266 should be adequate for 50-100 servers.
The box I'm running on is an old (7-8 years) Solaris server with a
900 MHz UltraSparc II processor. That's roughly comparable to a PII
running at 1.2 GHz. And it handles 25 times as many hosts as you are
aiming for.
> Third is plugins. Are BB plugins compatible with Hobbit?
Yes.
> Also how hard are plugins to write for Hobbit?
Plugins that run on the monitored client systems are as easy to write
as for BB, since it is basically the same thing.
Hobbit also allows you to write plugins for the Hobbit server, which
receive events from the Hobbit server daemon. This is used by the
core Hobbit tools - e.g. the hobbitd_rrd processes you see in the
ps-listing above are a plugin that handle updating of the RRD files
from the status- and data-messages that are sent to Hobbit. There
aren't any third-party plugins that use this yet (at least, I
don't know of any), but writing them is fairly simple since it
basically involves reading data from a pipe and processing it in
whatever way you want.
> I don't know if these even exist for
> bb, but I ultimately would like to integrate plugins that 1) monitor
> legato tape backup,
Dont know about this.
> 2) run nmap to see what ports are open/can be seen from an external
> machine,
The Hobbit client in version 4.2 (about to be released soon) reports
details about the network services running on a host. So you can check
for which ports are open/listening for connections, and trigger alerts
if any unwanted ports show up.
> 3) run 'lshw -html' to show a list of all the hardware on the system,
This would typically be a client-side test.
> 4) monitor uptime,
This is standard.
> 5) monitor OS and kernel versions (uname -a and head -n 1 /etc/issue),
This data is collected by the Hobbit client.
> 6) maybe some more router/network monitoring stuff and
Hobbit comes with built-in network service monitoring. There is also
an SNMP add-on which can be used for monitoring devices such as routers.
> Fourth is relay. By this I mean monitoring systems on a private
> subnetwork that are only accessible to the Hobbit server by going
> through an intermediate server. Is this possible with Hobbit and is it
> any more difficult to do than on BB?
Two ways of doing that. First, there is a proxy utility which is used
to forward Hobbit messages from one network to another. This is used if
your client systems on the private subnet are allowed to make outgoing
connections to the proxy, and the proxy can connect to the real Hobbit
server.
Second, Hobbit 4.2 includes a set of tools where it's the server that
contacts clients to pick up the data they have collected (i.e. the
traffic is initiated by the server, where the normal BB setup is for
the client to initiate the connection). Useful for DMZ style setups
where clients are not allowed to generate outbound connections.
> Fifth is portability. BB is very portable, I can make a 'model' client
> for say Red Hat and tar it and distribute it very easily to every
server
> I have using only a few commands. Is Hobbit the same, or are there
> client dependencies or other things that may make this more difficult.
The Hobbit client uses only the system libraries and standard utilities
found on your client systems. You will need at least one system where
you can compile the client binaries (that's similar to the BB
requirements), since a few of the client-side tools are written in C.
Once you have a client compiled for an OS, it is as portable as any
binary that is dynamically linked on your platform. I.e. you can
just copy it over as long as the same run-time libraries are available.
So far, we haven't managed to find any unix-like system that couldn't
run the Hobbit client. Including some rather odd ones. The current list
of client-side data collectors are
hobbitclient-aix.sh hobbitclient-darwin.sh hobbitclient-freebsd.sh
hobbitclient-hp-ux.sh hobbitclient-irix.sh hobbitclient-linux.sh
hobbitclient-netbsd.sh hobbitclient-openbsd.sh hobbitclient-osf1.sh
hobbitclient-sunos.sh
> Sixth is development. How active is the development of Hobbit, how big
> is the community, etc? How many people can attest to having fully
> functional hobbit setups, how long has it been around and how often
are
> new releases usually made?
Hobbit started back in late 2002 when it was called the "bbgen toolkit".
It was renamed to Hobbit in March 2005 when it had developed into a
complete replacement for BB. More details in the hobbit(7) man-page
available online at http://www.hswn.dk/hobbit/help/manpages/
It is actively being developed by me, but people on this list have
made contributions of code. Some have picked up special projects
like the Windows client and run that completely on their own.
I'd say Hobbit currently has a very active user community, and
the development community is slowly growing beyond just myself.
There are currently 433 subscribers to the Hobbit mailing list.
According to the Sourceforge download statistics, it is downloaded
about 1000 times per month.
http://sourceforge.net/project/stats/?group_id=128058&ugn=hobbitmon&type
=&mode=year
There was a thread on the mailing list back in May about who uses
Hobbit. The results were summarized here:
http://en.wikibooks.org/wiki/System_Monitoring_with_Hobbit/User_Guide#Wh
o_use_Hobbit_.3F
New releases have usually happened frequently - 2-4 times a year.
The current interval between the 4.1.2 release and version 4.2 is
unusually long - a whole year. I don't expect that to happen again.
> Also I saw something this morning about a Windows client -- how
> stable is that?
>From what I hear it should be usable. But you can stick with the
current BBNT client until it reaches version 1.0.
> How stable is the Solaris version?
Rock-solid.
> Is there a client for Mac OSX?
Yes. It will run the Hobbit server also, if you want to.
> Is Hobbit like BB in the sense that you can change paths to system
> binaries like grep and sed to allow easy use on other UNIXes like OSX?
Adding a client for a new OS will require implementing both a
client-side script to collect whatever data is interesting for this
system, and implementing the data parsing on the Hobbit server-side.
So it is somewhat more challenging. But since Hobbit already supports
all of the common Unix systems, I doubt that you will need to worry
about that. If you do have a system which is not on the list, I will
help you with adding support for it.
> When will 4.2 be officially released as a production version?
Probably by the end of this week.
> Since we have a working BB setup for now, I need to
> decide if I should try to start migrating now or if I should wait some
> time for Hobbit to develop more before I migrate from BB.
I don't think you have to wait. But it's for You to decide.
Regards,
Henrik
To unsubscribe from the hobbit list, send an e-mail to
hobbit-unsubscribe at hswn.dk
More information about the Xymon
mailing list