3rd interface not failing back…



  • I have a QMOE connection between 2 data centers, location A and B. Each d/c is running a pf setup of 2 machines, PF1 and PF2. Each machine has 4 int, 1-WAN, 2-LAN, 3-QMOE, 4-PFSync.
    Most traffic travels from A to B, not the other way. At some point the interfaces on location A failover from PF1 to PF2. Things seem fine. But when they failback to PF1, int-3 doesn't seem to failback, PF1 shows backup, PF2 shows master. But all traffic appears to be traveling to PF1 but the routes are not active (since it is in backup).
    I have added a screen shot to show how PF1 looks when it is not working.
    Thanks for any info.
    Jake



  • Check the switch.  It is not passing multicast correctly.

    Also search the forum.  This exact question has been asked (and answered) around 20 times.



  • We are looking into the multicasting. Sorry for duplicate post, but I searched for an hour looking for something similar and didn't find it.
    If you know of one of them please link to it here, thanks.

    Jake



  • log on with SSH or directly from console (choose 8 ).
    On the box not failling back, do a tcpdump like this:
    tcpdump -i <ifname>-ttt -n proto CARP
    where <ifname>is the name of the physical interface.

    Do the same on the second box.

    When all is working fine, you should see a trace showing multicast traffic directed to 224.0.0.18 (vrrp v2 multicast address), sourced from the IP address of the physical interface of the master node. On the slave node, you should see these packets too. When powering off  the master node the packets should then be sourced from the slave node with a higher advskew.

    The four main problems you should encounter:

    1. Misconfiguration: password, VHID or advskew problems, check it again.

    2. Another device using VRRPv2 is using a VHID you are using, check you network devices or change VHID

    3. You don't see master's packets on the slave node when doing the tcpdump (so the slave node has one or more interface in master mode). You have a communication error between the two machines. Check the switchs, the cables. Or look at problem 4 ;-)

    4. You have a NAT rule, natting everything from a source network to a single IP address which IS NOT the interface address and which is in ANOTHER subnet. Should happen on WAN iface most of the time.</ifname></ifname>



  • Stickying thread.



  • tcpdump -i xl0 -ttt -n proto CARP
    Here is the output of my tcpdump:
    709630 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    293069 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 002309 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    487570 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    514636 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 001317 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    267018 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    734179 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 001057 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    047719 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    953636 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    829337 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    171683 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 001111 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    610157 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    391038 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 234670 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    157247 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    1. 039601 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36

    the 151 is the master machine, the 251 is the machine on the other side of the QMOE link that is the other firewall PFsense box. you can see the vrid is different, so that shouldn't affect it.

    1. Misconfiguration: password, VHID or advskew problems, check it again.

    Checked this, it is correct.

    1. Another device using VRRPv2 is using a VHID you are using, check you network devices or change VHID

    Obviously it is connected to the pfsense on the other side of the qmoe, but not sure if vrid is same as vhid, but I manually checked in the gui for the config of both sides of qmoe and the vhid is different.

    1. You don't see master's packets on the slave node when doing the tcpdump (so the slave node has one or more interface in master mode). You have a communication error between the two machines. Check the switchs, the cables. Or look at problem 4 ;-)

    I see the master packets, see about tcpdump.

    1. You have a NAT rule, natting everything from a source network to a single IP address which IS NOT the interface address and which is in ANOTHER subnet. Should happen on WAN iface most of the time.

    Still checking this. But not sure what that would affect. Will post follow-up in a bit.

    thx for the help with this.



  • As far as NAT routing to single IP, we do not have that on the network that is having trouble.

    As you can see in the image the last rule is for the qmoe and it goes to * (all).




  • Can you give us a network diagram ? You have 4 machines as I can understand, 2 on A d/c and 2 on B d/c



  • Here is a simple drawing. The pf2 box, interface 4 (QMOE) is the only one that doesn't failback.




  • Are all of the nics the same type?



  • NIC types…
    pf1 and pf2:
    int 1 - Intel Pro 100 - WAN
    int 2 - Intel Pro 100 - LAN
    int 3 - Intel Pro 100 - pfsync
    int 4 - 3com 3C905-TX - QMOE

    pf251 and pf252:
    int 1 - Intel e1000 - LAN
    int 2 - Intel e1000 - WAN
    int 3 - Broadcom Gbit - QMOE
    int 4 - Broadcom Gbit - pfsync



  • Have you checked that either Foundry or HP equipment aren't filtering any type of trafic (like multicast)?



  • yep, multicast is working just fine. The foundry side is working fine… pf251,pf252. It is the HP side that is having the failback problem. But we checked the multicast and it is fine. I can also see it in the tcpdump on pf2.



  • Is there any other information you can give me? Anything else you might try? Please let me know.



  • Check network equipment on HP side.  Something is being blocked (multicast).



  • We are now on duplicate equipment as the other side. Foundry Super X. This did not solve the issue.



  • The equipment is not forwarding or blocking the CARP specific traffic.  Use tcpdump to monitor each machine to see if it is receiving the broadcast traffic.  I bet the switch is the culprit.



  • As you can see I did this already and the machines are seeing the carp traffic without any issue.

    @jakehathaway:

    tcpdump -i xl0 -ttt -n proto CARP
    Here is the output of my tcpdump:
    709630 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    293069 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 002309 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    487570 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    514636 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 001317 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    267018 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    734179 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 001057 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    047719 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    953636 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    829337 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    171683 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 001111 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    610157 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    391038 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    1. 234670 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36
    157247 IP 172.16.20.152 > 224.0.0.18: VRRPv2, Advertisement, vrid 6, prio 200, authtype none, intvl 1s, length 36
    1. 039601 IP 172.16.20.251 > 224.0.0.18: VRRPv2, Advertisement, vrid 5, prio 0, authtype none, intvl 1s, length 36

    the 151 is the master machine, the 251 is the machine on the other side of the QMOE link that is the other firewall PFsense box. you can see the vrid is different, so that shouldn't affect it.

    1. Misconfiguration: password, VHID or advskew problems, check it again.

    Checked this, it is correct.

    1. Another device using VRRPv2 is using a VHID you are using, check you network devices or change VHID

    Obviously it is connected to the pfsense on the other side of the qmoe, but not sure if vrid is same as vhid, but I manually checked in the gui for the config of both sides of qmoe and the vhid is different.

    1. You don't see master's packets on the slave node when doing the tcpdump (so the slave node has one or more interface in master mode). You have a communication error between the two machines. Check the switchs, the cables. Or look at problem 4 ;-)

    I see the master packets, see about tcpdump.

    1. You have a NAT rule, natting everything from a source network to a single IP address which IS NOT the interface address and which is in ANOTHER subnet. Should happen on WAN iface most of the time.

    Still checking this. But not sure what that would affect. Will post follow-up in a bit.

    thx for the help with this.



  • Well then about the only thing I can think of is the NICS in the machine.

    BTW: I have major problems with Broadcom nics + CARP at work.  It is a driver issue of some sorts.



  • So I have completely by-passed routing on the pf box since it isn't working. It works until it gets into the following state.  (see attached pics).






  • Iv seen this one before… sorry to say that im a noob and just figuring it out my own probs at:
    http://forum.pfsense.org/index.php/topic,10458.0.html

    At my configuration... it happened when the CARP suddenly "worked" after i sorted out some bugs... then again it didnt work. It was when the SYNC interfaces were on 10Mb/s old NICs. And the LAN VIP became master on Backup, WAN and WAN2 were left Master at the Master box. And then when i went to 100/10 NIC's the backup took all the VIP's as master... so it might be something different than your prob.

    One question... how would i bypass the "broadcast" thing if it really is the switch or NIC's bad appetite for not eating broadcast packets. ?


Log in to reply