Scaling

hello

15 000 devices here. For me the key is ssd :)

I plan to monitore 60 000 devices with xymon. Only network devices.

We'll see the result.

oau

Le vendredi 05 avril 2013 à 13:54 -0500, Larry Barber a écrit :

Well, I'm monitoring ~2000 hosts on a fairly modest box (8 3Ghz cores,
8 GB of memory). I'm also running quite a few cpu intensive scripts on
the same box that could be easily moved to another host, if needed. I
do the network testing on separate hosts in each of our major security
zones, for reliability of the tests more than to unload the main Xymon
server. The main server is not operating anywhere near its capacity,
it's using less than 10% (physical) of it's memory and the load
average tends to stay around 1. I suspect that the box could handle
5000 hosts without too much trouble, maybe more. 

If you do have scaling problems there are some things you can do,
though. Move things like the network tests to separate hosts. You can
also move the alerting to a different host using xymonproxy. I've
found that the most likely limit you're likely to hit with Xymon is
disk i/o, this can be helped by moving the data directory to SAN. 

Thanks,
Larry Barber


On Fri, Apr 5, 2013 at 12:57 PM, Bruce Ferrell <user-24fbf1912cfe@xymon.invalid>
wrote:
        Hi all,
                I've been doing systems monitoring for a very long time now...
        I was early on with BB, used HP openview back in day day, blah
        blah.
                Anyway, recently I've been told that in very large
        installations (multi thousands of devices) things like zabbix
        are the only thing(s) that will do.
                What are the groups thoughts on this?  What ARE the scaling
        limits of xymon and can they be overcome somehow?