Page 1 of 1

Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Fri Aug 10, 2007 1:24 pm
by believewireless
We have a problem with BGP on 2.9.42 where BGP just sort of locks up. Running winbox into the public address of the router fixes it. Is there a way to see if you've lost a connection with a peer and then reboot the server?

Basically, the only thing I don't know how to do is determine that BGP has been down for a certain period of time and remember that in a script. Any help would be useful.

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Fri Aug 10, 2007 7:29 pm
by changeip
i would try to find the source of the problem rather than bandaiding it... just my opinion. anyhow, can you take a supout during this 'disconnect' period and send to support ? Also, seems like it might be a memory leak possibly as that's something they've worked on in the past 2-3 releases ?

Sam

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Fri Aug 10, 2007 7:40 pm
by believewireless
A memory leak is my guess as well. Always seems to happen at 4 or 5 AM though. The problem has been worse with .43 and .44. We haven't been brave enough to try .45 since we usually had lockup problems on the other two versions.

So, for now, we'd rather just write a script that fixes the problem until Mikrotik gets a working version.

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Fri Aug 10, 2007 8:08 pm
by changeip
ive been waiting for a stable version since .26 : ) I don't dare upgrade those since they are just working with limited numbered routes.

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Mon Aug 13, 2007 10:15 am
by ste
ive been waiting for a stable version since .26 : ) I don't dare upgrade those since they are just working with limited numbered routes.
Hi,

I started using ROS for a BGP Backuplink for testing after .26. I started with .35.
I checked each release since then. So I never had a stable BGP backup router
with ROS. Now I give 2.9.45 a try (routing-test).
I dont think it's a memory problem. The router (it's a vmware machine) runs
very stable. OSPF had no issues at all. The machine has a enough memory
and cpu is idling.
The behavior is that the BGP connections work between 3 and 6 days then
they restart or they hang. When I login with winbox the BGP hanging
connections start working again without any action than loggin in.
The effect is on all connections eBGP and iBGP.

I'm wondering how people could live with this as their main BGP Router.
Or do all stick on 2.9.26?

Stefan

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Mon Aug 13, 2007 7:25 pm
by changeip
this problem of logging in with winbox to fix it happened to me in certain versions. I don't know why. Simply logging in would fix it. I think the probing of the routing table triggered it or something. Anyhow, try 2.9.45, if it's still an issue please make a supout and email it to support... without a supout on the latest version these types of problems will never get fixed and we'll all suffer thru 2.9.99 until we get stable BGP. I cannot stress how important it is to actually report these issues and not just post in the forums about them.

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Mon Aug 13, 2007 7:48 pm
by believewireless
I asked for a script, not to argue about BGP. Does anyone have a script that can see if no routes have been there for five minutes and then reboot the server?

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Mon Aug 13, 2007 9:05 pm
by changeip
ill help you write this. would it better to check if idle time > 5 mins ? Let me dig into the scripting and see what i can come up with

Sam

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Fri Aug 24, 2007 7:27 am
by ste
Hi,

2.9.45 routing-test connections are now up for 10 days. Never reached
this with other versions. So it seems to be better now.

Stefan

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Mon Aug 27, 2007 4:46 pm
by ste
A memory leak is my guess as well. Always seems to happen at 4 or 5 AM though. The problem has been worse with .43 and .44. We haven't been brave enough to try .45 since we usually had lockup problems on the other two versions.

So, for now, we'd rather just write a script that fixes the problem until Mikrotik gets a working version.
So give .45 a try. I've running BGP-Sessions now for 2 Weeks on our Backup-BGP Router
with .45.

Stefan

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Fri Nov 30, 2007 9:14 pm
by airtech
I am running .49 and can almost bet money on the fact that every 5-6 days my BGP will stop routing our traffic out and the only thing that fixes it is a reboot of the router. I unfortunately did not try .45, went directly from .44 to .49 just to find that I have the same problem. I am running the routing-test package still to no avail. What is weird is I still have an established BGP connection, still have a default route, it just won't route the traffic. Next time this happens I will try and get a support.rif file, it just always happens at the worst of times when I need to get it back so fast that I just reboot it. I am beginning to wonder if the BGP problems will ever get fixed.

Keith

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Sat Dec 01, 2007 12:04 am
by ste
Hi,

.45 and .46 runs fine here. .45 virtual on a vmware server on a P4 2,8
and .36 on a P4 1,6. All routing-test. All machines have OSPF and
Vlans running.

Stefan

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Sat Dec 01, 2007 12:32 am
by airtech
Can you upgrade one of them to .49 and see if they still work for you? I would be very interested to see if they do or don't. If they don't I will just downgrade to .46 until there is a true resolution.

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Sat Dec 01, 2007 9:10 am
by ste
Can you upgrade one of them to .49 and see if they still work for you? I would be very interested to see if they do or don't. If they don't I will just downgrade to .46 until there is a true resolution.
:shock: No. I do not play with my bgp-routers without strong need.

Stefan

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Sun Dec 02, 2007 6:38 am
by airtech
I figured, but I had to ask. :D

Re: Reboot Server if BGP Routes Lost for More than 5 Minutes

Posted: Mon Dec 10, 2007 2:31 pm
by bokili
Can somebody of you, send packages for 2.9.45 or 2.9.46 ?