Re: OpenBGP Issues. :-(
Alex Mathiasen(a...@mira.dk) on 2013.02.28 14:51:25 +0100: Dear recipients, I have been using OpenBGP for a while with OpenBSD - And I am very satisfied with the performance and amazed by the ease of configuration. My BGPD is configured against a Danish ISP called TDC - And we were previously configured to receive a full routing table. However a few months ago I ran into an issue where my BGPD stopped working properly. Was this in November by any chance? [ Alex Mathiasen ] Yes, it was at 29.11.2012. Happened in the middle of the night.. :-( It appeared the BGPD kept receiving the routing tables, and then start all over. Looking into the log files, it appeared BGPD received a certain route in the routing table, and then grumbled about the prefix, apparently for some reason the result was BGPD kept reloading when it reached this route. The result was of course my network was down. As TDC (My ISP) couldn't resolve which route that caused this issue (They told me: That's what happened when you use third party software, so no help there...), we agreed that my connection would be set to Default candidate, instead of receiving a full routing table. So now I have configured a static route to forward all my traffic to this route. However this is not the result I wanted, as I am about to have one more connection, so I have 2 connections outbound. But the automatic failover switch / load balancing won't work, as long as I have my static route. This is why I want to go back to receiving a full routing table. Is there any way of configuring BGPD to ignore a specific route in case of corrupted prefix, so this won't happened again? No there is not such a feature, and the bgp protocol mandates session teardown in certain cases anyway. Your report lacks a few details, please send with dmesg next time. And your bgpd.conf is not valid. [ Alex Mathiasen ] I do apologize for the lack of information, I was unable to find my logfile from that date, and was unable to provide with more information. My guess is that your problem is fixed by the patch available on http://www.openbsd.org/errata52.html [ Alex Mathiasen ] It would appear this is the patch I need to resolve this issue. So I will try to apply this patch, thank you! You could also update to -current. /Benno
OpenBGP Issues. :-(
Dear recipients, I have been using OpenBGP for a while with OpenBSD - And I am very satisfied with the performance and amazed by the ease of configuration. My BGPD is configured against a Danish ISP called TDC - And we were previously configured to receive a full routing table. However a few months ago I ran into an issue where my BGPD stopped working properly. It appeared the BGPD kept receiving the routing tables, and then start all over. Looking into the log files, it appeared BGPD received a certain route in the routing table, and then grumbled about the prefix, apparently for some reason the result was BGPD kept reloading when it reached this route. The result was of course my network was down. As TDC (My ISP) couldn't resolve which route that caused this issue (They told me: That's what happened when you use third party software, so no help there...), we agreed that my connection would be set to Default candidate, instead of receiving a full routing table. So now I have configured a static route to forward all my traffic to this route. However this is not the result I wanted, as I am about to have one more connection, so I have 2 connections outbound. But the automatic failover switch / load balancing won't work, as long as I have my static route. This is why I want to go back to receiving a full routing table. Is there any way of configuring BGPD to ignore a specific route in case of corrupted prefix, so this won't happened again? I hope that some of you have an answer for this... Here you can see my bgpd.conf: AS router-id 000.000.000.000 network 000.000.000.00/00 neighbor 000.000.000.000 { remote-as descr TDC local-address 000.000.000.000 passive holdtime180 holdtime min3 tcp md5sig password 00 } log updates
Re: OpenBGP Issues. :-(
Alex Mathiasen [a...@mira.dk] wrote: It appeared the BGPD kept receiving the routing tables, and then start all over. You don't mention which version of openbsd you are using. There are some problems like this in older versions of bgpd which are now fixed. You may want to try a new snapshot.
Re: OpenBGP Issues. :-(
Alex Mathiasen(a...@mira.dk) on 2013.02.28 14:51:25 +0100: Dear recipients, I have been using OpenBGP for a while with OpenBSD - And I am very satisfied with the performance and amazed by the ease of configuration. My BGPD is configured against a Danish ISP called TDC - And we were previously configured to receive a full routing table. However a few months ago I ran into an issue where my BGPD stopped working properly. Was this in November by any chance? It appeared the BGPD kept receiving the routing tables, and then start all over. Looking into the log files, it appeared BGPD received a certain route in the routing table, and then grumbled about the prefix, apparently for some reason the result was BGPD kept reloading when it reached this route. The result was of course my network was down. As TDC (My ISP) couldn't resolve which route that caused this issue (They told me: That's what happened when you use third party software, so no help there...), we agreed that my connection would be set to Default candidate, instead of receiving a full routing table. So now I have configured a static route to forward all my traffic to this route. However this is not the result I wanted, as I am about to have one more connection, so I have 2 connections outbound. But the automatic failover switch / load balancing won't work, as long as I have my static route. This is why I want to go back to receiving a full routing table. Is there any way of configuring BGPD to ignore a specific route in case of corrupted prefix, so this won't happened again? No there is not such a feature, and the bgp protocol mandates session teardown in certain cases anyway. Your report lacks a few details, please send with dmesg next time. And your bgpd.conf is not valid. My guess is that your problem is fixed by the patch available on http://www.openbsd.org/errata52.html You could also update to -current. /Benno
Re: OpenBGP Issues. :-(
On 2013-02-28, Alex Mathiasen a...@mira.dk wrote: Looking into the log files, it appeared BGPD received a certain route in the routing table, and then grumbled about the prefix grumbled about is not very exact, actual log entries would be a lot more helpful. It would be even better if you could capture the actual update messages causing the problem (tcpdump -i em0 -w bgp.pcap -s 1500 port 179 and host $foo) As TDC (My ISP) couldn't resolve which route that caused this issue (They told me: That's what happened when you use third party software, so no help there...), Every BGP implementation has problems from time to time, IMHO anyone running this really needs to keep track of development of their chosen implementation/s (at least keep an eye on changelogs / cvs commits / mailing lists etc) and general network problems (nanog, local network operator groups, etc), and when they do have problems provide good information to the (vendor | developers | 3rd party support org). Also see everything that Benno wrote. :)
Re: OpenBGP Issues. :-(
Alex Mathiasen a...@mira.dk a écrit : Dear recipients, I have been using OpenBGP for a while with OpenBSD - And I am very satisfied with the performance and amazed by the ease of configuration. My BGPD is configured against a Danish ISP called TDC - And we were previously configured to receive a full routing table. However a few months ago I ran into an issue where my BGPD stopped working properly. It appeared the BGPD kept receiving the routing tables, and then start all over. Looking into the log files, it appeared BGPD received a certain route in the routing table, and then grumbled about the prefix, apparently for some reason the result was BGPD kept reloading when it reached this route. The result was of course my network was down. As TDC (My ISP) couldn't resolve which route that caused this issue (They told me: That's what happened when you use third party software, so no help there...), we agreed that my connection would be set to Default candidate, instead of receiving a full routing table. So now I have configured a static route to forward all my traffic to this route. However this is not the result I wanted, as I am about to have one more connection, so I have 2 connections outbound. But the automatic failover switch / load balancing won't work, as long as I have my static route. This is why I want to go back to receiving a full routing table. Is there any way of configuring BGPD to ignore a specific route in case of corrupted prefix, so this won't happened again? I hope that some of you have an answer for this... Here you can see my bgpd.conf: AS router-id 000.000.000.000 network 000.000.000.00/00 neighbor 000.000.000.000 { remote-as descr TDC local-address 000.000.000.000 passive holdtime180 holdtime min3 tcp md5sig password 00 } log updates Hi, Please have a look in archives for a similar thread i did initiate.