RC1 CPU overload (dropped tcp sessions)

RouterOS General
1

I use hotspot and universal nat (or however it calls now). Upon there were 2 clients - everything was ok. Now there is almost 100 clients and from time to time (sometimes few minutes, sometimes about hour) all sessions are dropped (while pinging I get “unreachable network”) for a moment. Is it my mistake or bug?

kj

2

hmm… I found that the problem occours when CPU load is 100%… So question is why cpu load was today 100% 2 times (even couple of hours, only reboot fixes it), when after reboot the load is between 5 and 10 %? anyone see the same?

3

now I have 100% again - after 15 minuts of working without changeing anything. 25 users online. CPU - Duron 700, 512M RAM, bandwidth 2x2Mbit max (current real bandwidth about 2x512k), no wireless cards.

kj

4

Take a look at the connections in the firewll the problem could be a ddos attack or something if you see hundreds of connectiions open? But thinking back I had to use a 1500 mhz processor with around that many users with PCQ queues and alot of simple queues because of a simmilar effect.

5

I can see hundreds of connections - most of them from p2p. Do you think it’s the cause?

6

Yup. Try to limit overall connection count through the router.

7

Use:

connection-limit

in firewall filters so each user can only get like 60 connection each.

8

thanks. good idea - but what about an udp connections? this limit applies only to tcp connections…
kj

9

You are right. for IPTables on linux there is a patch available to do a udp connlimit but you can’t do it with Mikrotik. I did post asking if there would be such a feature a while back but got no feedback. I guess it would be pretty simple for them to add it.

Still in practice limiting TCP connections seems to be pretty effective I think the TCP connections do alot of management so the P2P eases off when it realizes the limit.

10

so:

  1. replaced cpu (athlon 2000+)
  2. limited tcp connections (but still lot of them, most are udp), but native linux in the same network on PII/400 run ok - I think router is for connecting hosts and routing packets :slight_smile:

cpu load in “normal” operations is around 5%, then (sometimes after 5 minutes, sometimes 2 hours) goes to equal 100% and no lower until reboot
sessions are still dropped ocassionaly - I see no rule, when?

I’m not sure, if one thing run on second, or these are two different problems? but I’m really starting worry about it. anyone got the same???

system resource> pr
uptime: 1h1m9s
free-memory: 483624kB
total-memory: 516796kB
cpu: “AMD”
cpu-frequency: 1668MHz
cpu-load: 100
free-hdd-space: 29438kB
total-hdd-space: 61596kB
write-sect-since-reboot: 3124
write-sect-total: 807538

and one more - sometimes router reboots himself
my config is pretty long, so I can put any fragment you want, but not whole

kj

11

Send support-output file from the router to support@mikrotik.com