Re: RPMs for spread/wackamole/etc and wackamole issues

Spread+wackamole is BLAH! Use UCARP. It works. :o)

On Nov 9, 2007 4:31 PM, Spangler, Tim <[email protected]> wrote:
>
> I hate to follow-up on a two month old problem, but I'm now seeing something
> similar.
>
> I've been running several wackamole installs in production since march, and
> recently came across following line in my messages file: wackamole[8532]
> "953 No such interface", which appears to be similar to the problem in this
> message.
>
> Additionally, I'm running 2 wackamole clusters in the same subnet, and have
> recently begun seeing the following errors on one of the sets of clusters:
> In_my_component: unknown proc 192.168.252.25.  This IP address is from a
> machine in the other cluster, but spread is running on a different port, so
> as far as I can tell, they shouldn't be able to see each other.
>
> Any thoughts out there?
>
> Thanks,
>
> -Tim Spangler
>
>
> ________________________________
> From: Jordan Mendler [mailto:[email protected]]
> Sent: Thursday, September 13, 2007 4:26 PM
> To: [email protected]; [email protected]
> Subject: [Wackamole-users] RPMs for spread/wackamole/etc and wackamole
> issues
>
> Hi all,
>
> I want to preface by letting everyone know that I have built some RPMs for
> spread, wackamole, mod_log_spread and spreadlogd. They are available from
> http://biopackages.net/ for CentOS4 and some Fedora distributions. They will
> eventually be built for every CentOS and Fedora release when we expand our
> repository. And we have some updated RPMs in our testing repository that
> should be pushed out in a few days. The SRPMs are also available if you
> would like to build your own.
>
>
> Now onto the problem...
>
> In configuring Wackamole, I have been having some issues, so I am hoping
> that someone will be able to help me get this working. My setup is CentOS4
> x86_64 (linux 2.6.x) if that matters. Wackamole/spread builds and installs
> fine and spread works fine. Wackamole starts fine and seems to think it
> works, but in reality it does not. Therefore I am not sure if this is an
> issue with Wackamole's interaction with OS, or what. I have been following
> Theo's "Scalable Internet Architectures" book in an attempt to get wackamole
> setup.  This is also in a testing environment ( 10.x.x.x IP's), before we
> tried it with real IPs and had the same issue.
>
> What happens:
> -Start spread on both systems, everything works fine. It is configured to
> use 10.67.183.121 and .122, respectively, which are both setup on eth1 and
> independent of the wackamole IPs. The same thing happens if we setup spread
> on the same IP as wackamole.
> -Start wackamole on both systems and:
>    (1) eth0 which is configured with 10.67.183.116 on one and 10.67.183.117
> on the other is taken down by wackamole such that an 'ifconfig' only shows
> eth1 up
>    (2) wackatrl -l appears to be working properly, showing the following on
> each system:
>    Owner: 10.67.183.116
>            *    eth0:10.67.183.116/32
>    Owner: 10.67.183.117
>            *    eth0: 10.67.183.117/32
>
> Despite #2, neither machine brings up .116 or .117. There is obviously
> something going on, because from another machine I can still ping/ssh into
> 116 and 117, which may be as a result of arp.
>
> At this point if I kill spread on one of the 2 machines (say the one with
> .117), wackatrl -l shows what appears to be correct:
> Owner: 10.67.183.116
>         *    eth0: 10.67.183.116/32
> Owner: 10.67.183.116
>         *    eth0:10.67.183.117/32
>
> Despite what wackatrl thinks, I am now able to ping/ssh into only one of the
> IPs, and the IP of the machine that was taken down is not brought up on the
> other machine. The whole thing seems to be acting weird.
>
>
> The only indication I can find is /var/log/messages which shows the
> following on both machines. I am not sure if this is a 2.6 kernel not
> supported issue (hopefully not, cause I would really like to get wackamole
> working):
> Sep 13 07:44:50 JMM1 wackamole[26151]: connecting to 4803
> Sep 13 07:44:50 JMM1 wackamole: wackamole startup succeeded
> Sep 13 07:44:50 JMM1 wackamole[26151]: DOWN: eth0:
> 10.67.183.116/255.255.255.0
> Sep 13 07:44:50 JMM1 wackamole[26151]: 953 No such interface
> Sep 13 07:45:02 JMM1 wackamole[26151]: 911 No such interface
> Sep 13 07:45:02 JMM1 wackamole[26151]: Re-queued arp spoof notifier for
> virtual entry.
>
> Also, when I try a wackamole.conf with 4 IPs, wackatrl shows:
>    Owner: 10.67.183.116
>            *    eth0:10.67.183.116/32
>    Owner: 10.67.183.117
>            *    eth0:10.67.183.117/32
>    Owner: 10.67.183.124
>            *    eth0:10.67.183.124/32
>    Owner: 10.67.183.125
>            *    eth0:10.67.183.125/32
>
>
> My configurations look as follows (it is the same on both machines), though
> I have tried many other configurations such as changing /32 to /24, trying
> other IPs and so on:
> [root@JMM2 etc]# cat /etc/wackamole.conf
> Spread = 4803
> SpreadRetryInterval = 5s
> Group = wack1
> Control = /var/run/wack.it
>
> Prefer None
>
> VirtualInterfaces {
>         { eth0:10.67.183.116/32 }
>         { eth0: 10.67.183.117/32 }
> }
>
> Arp-Cache = 10s
> mature = 5s
>
> Notify {
>         eth0:10.67.183.1/32
>         arp-cache
> }
>
> Balance {
>         AcquisitionsPerRound = all
>         interval = 4s
> }
>
> The importand part of Spead.conf is:
> Spread_Segment  10.255.255.255:4803 {
>         JMM1    10.67.183.121
>         JMM2    10.67.183.122
> }
>
> Any assistance is greatly appreciated.
>
> Thanks so much,
> Jordan
>
> _______________________________________________
> wackamole-users mailing list
> [email protected]
> http://lists.backhand.org/mailman/listinfo/wackamole-users
>
>