Re: Network problem (Connection refused) with 2.0.33?

Chris Adams (cadams@ro.com)
15 Jan 1998 05:01:00 GMT

According to Chris Adams <cadams@ro.com>:
>Once upon a time, Alan Cox wrote
>> > Jan 14 15:59:23 sh1 sendmail[832]: NOQUEUE: SYSERR(root):
>getrequests: accept: No route to host
>>
>> > after the connections are refused sometimes (even when connecting from
>> > the same system). Also, the monitoring program that checks dns, web,
>>
>> Default sendmail behaviour is to sleep a few seconds after an accept failure.
>> Dont ask me why - run exim or take the silly sendmail thing out (I thought
>> it was gone in 8.8.7 anyway)
>
>I just found that line in sendmail (8.8.8): daemon.c line 335 does a
>"sleep(5)" after accept fails.
>
>I can fix that, but why does it fail in the first place? And what about
>FTP? I don't think inetd has the same behavoir as sendmail, and why
>would they fail at the same time (but nothing else on the system)?
>Could another program be interfering with them?

I hate to follow up to my own post, but I have a little more data (a
several more questions):

I recompiled sendmail with the sleep(5) after an accept failure taken
out. This has solved the problem for practical purposes. We still have
tons of "accept: No route to host" messages being logged (on average one
every six seconds) though. For you networking gurus: under what
conditions would accept return EHOSTUNREACH? It doesn't return until
the TCP handshake is completed, right? How could a host then be
unreachable? Is there any way to track this? It looks like the addr
field is not changed when an error occurs - can I find the address
somewhere?

-- 
Chris Adams - cadams@ro.com
System Administrator - Renaissance Internet Services
I don't speak for anybody but myself - that's enough trouble.