Hello,
Our apologies to Martin and Benoit for beeing to late with our fix.. We first had to consider the options we have and find a solution which works for us and the "victims".
Martin Blapp wrote:
Hi all,
The recent bluewin migration to new dns load-balancers broke sendmail on BSD systems. It seems that the BSD resolver (not relevant if one uses bind8 or bind9) is different than the one in linux.
it indeed is. To be honest, the Load Balancers do "the wrong thing"
As a workaround on our side we now implemented a DNS change: - Instead of only one mx entry for bluewin.ch we now have three
; "load balanced entry" - delegated to LB's bluewin.ch MX 10 mxbw.bluewin.ch bluewin.ch MX 42 mxiab.bluewin.ch ; local DNS entry to LB 1 bluewin.ch MX 42 mxzhh.bluewin.ch ; local DNS entry to LB 1
Unpatched sendmail will fail on the first MX entry and fall back to the MX with the lower priorities.
I suspect that all BSD systems have this problem, including MacOSX.
The following sendmail patch does work around this problem. It is not 100% correct but does work for now.
--- sendmail/daemon.c.orig Thu Mar 17 10:15:11 2005 +++ sendmail/daemon.c Thu Mar 17 10:12:34 2005 @@ -2256,7 +2256,7 @@ /* check for name server timeouts */ # if NETINET6 if (WorkAroundBrokenAAAA && family == AF_INET6 &&
errno == ETIMEDOUT)
(errno == ETIMEDOUT || h_errno == TRY_AGAIN)) {and of course you'll need this option in your MC File:
define(`confBIND_OPTS', `WorkAroundBrokenAAAA')dnl
The patch is useful nevertheless. Other ISP's might not recognize the problem or not fix it on their side...
Guido Roeskens Bluewin