Re: Re: hmm, little more help continued...
Theo Schlossnagle <[email protected]> Thu, 15 Apr 2004 19:11:12 -0400
| Newsgroups | gmane.comp.apache.mod-wackamole.general |
|---|---|
| Message-ID | <[email protected]> |
On Apr 15, 2004, at 3:29 PM, Superman wrote: > hmm, I'm still hoping someone can shed some light on this. Today for > the > first time ever the cvs version of wackamole started and binded an ip > address on the interface. pretty much the same configuration as > before but > this time before I started wackamole I cleared the arp table. Is there > something that maybe hangs wackamole startup if the arp-table is too > big?? > There were 62 arp entries in the arp table when I originally tried to > start > wackamole. (67 according to wackamole below). After clearing out the > arp > tables with a script wackamole started and binded the virtual ip. I > am > still trying to resolve this problem, as I really think this is a great > product. Thank you all for your help, and suggestions. The ARP sampling that happens on Sun is black magic. I wrote it and I don't know how it works. Which likely means that it doesn't work. There is no sound documented way of retrieving the local arp cache from a Solaris machine. So, in arpcache.c, there is a section that says: #elif defined(DL_UDERROR_IND) blackmagic #else Try changing that to #elif 0 which will completely disable the Solaris arp cache support -- but the rest of wackamole should work fine. Additionally, if you have platinum support and can get your hands on a Sun software engineer, I'd love to know the "right way" of fetching the arp cache from a Solaris box without invoking an outside program -- like "arp" or "netstat". The arp tables dumped in the output of your last mail look quite sound. So, apparently my hackish attempt to reverse engineer the ioctls/putmsg/getmsg blackmagic that Solaris' netstat performs on /dev/ip. I asked someone who has the Solaris source and they said that the guts of that were oddly missing our of their source tree... So, who knows. You're also more than welcome to muck with that section of C, perhaps one of the buffers is too small -- just don't look for rhyme or reason, it was written based of the output of truss and a hex editor :-) // Theo Schlossnagle // Principal Engineer -- http://www.omniti.com/~jesus/ // Postal Engine -- http://www.postalengine.com/ // Ecelerity: fastest MTA on Earth