[Xymon] Most number of hosts

[Henrik Størner]

On 22-05-2012 19:00, Michael Beatty wrote:
> Is there any close proximity number available as to how many clients
> Xymon can handle?  I've seen in the wiki pages that one user is
> advertising over 10,000 however, that seems to be in a "clustered"
> setting.  I see a few more users suggesting over 2,000.  My concern lies
> in that my plans stand at some where in the middle of that, and also,
> there really is no gauge as to what hardware and network topology those
> systems are running.
>
> My environment will be in the ball park of about 5,000 systems and many
> of them are on satellite bands with some latency concerns, pings tests
> come back at around 3 seconds.

I don't have much experience with high-latency connections - you'll 
obviously have to tweak some network timeout settings.

The main performance issue in Xymon these days are the disk I/O from 
updating all of the RRD files. Here are some numbers fresh off my main 
server:

$ iostat
Linux 2.6.32-5-amd64 (myserver)       05/22/2012      _x86_64_        (8 
CPU)

avg-cpu:  %user   %nice %system %iowait  %steal   %idle
            1.95    0.06    0.79    0.65    0.00   96.56

Device:            tps   Blk_read/s   Blk_wrtn/s   Blk_read   Blk_wrtn
cciss/c0d0        5.81        40.83       142.54   23238617   81132930
cciss/c0d1       79.20        31.19      3220.84   17755352 1833330720

It's been up for close to a week. Both of these disks are RAID-1 
devices; c0d0 are normal 10K RPM disks, c0d1 are SSD disks and holds 
ONLY the xymon/rrd and xymon/tmp directories. As you can see, the write 
activity on the SSD disks is huge - they are about a year old now, and I 
guess I'll have to replace them soon.

CPU- and memory-wise, this box is doing nothing at all. It has 2 
quad-core Xeon's and 24 GB RAM, but it only uses about 6 GB - the rest 
is used for disk cache (the way Linux does automatically).

Since the box has so much RAM and the RRD files only take up about 8 GB 
right now, I have been thinking about moving them to a RAM disk, and 
just rsync'ing them to the SSD disks once an hour or so.

It handles about 4000 full clients, and another 2500 for which we only 
do network tests.


Regards,
Henrik