sg-1100 DHCP lease status page load time

stephenw10

22MB seems huge for 100 clients. I'm not surprised it's slow opening that.

Have you looked inside that file? Hard to imagine it can all be legit data.

You might try just stopping the dhcpd service and removing that file then restarting it.

Steve

Gertjan

@gabric098
During high-season I've seen some 100 clients connection to 'my' pfSense every day.
Some 40 company devices, 60 or so from clients on another LAN : a captive portal.

-rw-r--r--  1 dhcpd  _dhcp  78167 Feb 22 09:57 dhcpd.leases
-rw-r--r--  1 dhcpd  _dhcp  79183 Feb 22 09:08 dhcpd.leases~

these "1592644918" numbers tell me that the DHCPD is trying to re manage the leases file, but it was interrupted while doing so.
Keep in mind : it's not only the number of DHCP client that determine the load. There can be devices that ask for a new lease every xx seconds or so (yes, these exists and should be banned from any network). That device will stress the DHCPD server and not let him doing what it should be doing.

Also, and this is completely IMHO : I wouldn't try to 'drive' a 100+ network with a "1100", depending your using packages that can stress the device.

gabric098

Hi @gertjan ,
thanks for your reply.
Please note that 80% of the time the network has a lot less clients than 100. I was trying to consider the most extreme situation that rarely happens. It's a home network with 2 families connected and a bunch of IOT devices.
I do also suspect something goes wrong when dhcpd tries to "rotate" the dhcpd.leases file.
I can definitely try to delete the file and restart the service but I'd like to understand what's going wrong to address the root cause of the problem.
What log file shall I look at for dhcpd logs?

Thanks,
Gab

Gertjan

@gabric098 said in sg-1100 DHCP lease status page load time:

What log file shall I look at for dhcpd logs?

Dono if the dhcpd has option to tell you what it's doing.
I advise you to :
Ditch de leases file.
Get rid of the "ridiculous" /16

A typical lease record, like :

lease 192.168.2.80 {
  starts 6 2021/01/16 08:06:31;
  ends 6 2021/01/16 14:06:31;
  tstp 6 2021/01/16 14:06:31;
  cltt 6 2021/01/16 08:06:31;
  binding state free;
  hardware ethernet ec:c4:0d:xx:e4:89;
}

is 200 bytes in size.

When they all get used, over you 3 networks of /16 or 65535 max entries per networks make 3 x 65535 * 200 = 39 Mbytes (worst case). You are half way, and the 1100 is already struggling.
Make your networks more classic, like /24, and the problem will auto solve.
Btw : dono if the actual pool size might influence all this.

gabric098

I've cleaned up the /var/dhcpd/var/db directory and manually rotated the leases file. Essentially I've renamed the dhcpd.leasesto dhcpd.leases~ and created an empty dhcpd.leases file.
Now the GUI loads in no time. I'll keep monitoring the directory as I believe something fails during what is described in the man page:

In order to prevent the lease database from growing without bound, the file is rewritten from time to time. First, a temporary lease database is created and all known leases are dumped to it. Then, the old lease database is renamed /var/lib/dhcpd/dhcpd.leases~. Finally, the newly written lease database is moved into place.

Thanks,
Gab

stephenw10

Ah, yes what was the scope size? If it was close to the full /16 and you have a lot of new devices I could imagine that file becoming huge.

Steve

gabric098

@stephenw10
I'm not sure what you mean with scope size but I believe to understand that the problem here is not the number of active leases but rather the fact that in that files there were loads of very old leases like the one I posted which was from may 2020 (expired in the same day).

gabric098

Ok, this is weird.
As I mentioned before I had created an empty dhcp.leases file this morning creating a "backup" version dhcp.leases~.
After few hours I find again the 21MB dhcp.leases files with the first entry:

lease 10.40.44.246 {
  starts 2 2020/04/14 20:42:23;
  ends 2 2020/04/14 20:52:50;
  tstp 2 2020/04/14 20:52:50;
  cltt 2 2020/04/14 20:42:23;
  binding state free;
  hardware ethernet 2c:56:dc:b7:b7:57;
  uid "\001,V\334\267\267W";
}

... april 2020 ?

Gertjan

I've a theory : a zero byte dhcpd.leases file isn't .. valid.
So it took the backup - it recognized it as valid, and rebuild a good, huge dhcpd.leases.

Rename your backup to anything except dhcp.leases~ ( name it back.old.do-not-touch)
Jtst delete dhcpd.leases, as dhcpd will create it anyway.

stephenw10

By scope size I mean what is the range?

You have /16 subnets there but I assumed you didn't a DHCP range with 65534 possible leases.
If you have a large number of new devices they will just keep claiming new leases without overwriting to old ones > massive file!

Steve

pete35

@gabric098

is it possible for you to change the scopes to
10.20.0.0/21 / 10.30.0.0/21 / 10.40.0.0/21 ?
This will give more than 2000 addresses per scope, should be enough for your system und the leases db will not grow that much anymore.

gabric098

I see what you mean @stephenw10 and @pete35 .
I've just reduced the DHCP range from 10.x.5.1 to 10.x.15.255 for all the VLANs. That will give me plenty of addresses without bloating the file.
The dhcpd.leases file immediately reduced its size to a more manageable 900KB, I'll keep monitoring it in the next days to see if it stays the same or keeps growing.
I'll update the thread.

Thank you all for your precious support.
Gab