Earlier, I posted the following message to the list:
Admin Staff wrote:
> It appears I have an ongoing problem that I think is related to sendmail.
> Occaisionally, it stops accepting new netconnections for a breif period of
> time (about 40 seconds) and then all of a sudden will pass them through.
> I believe this has to do with sendmail, because I notice it often when I
> am sending an email in the latest version of Pine (it calls sendmail in
> the foreground). What will happen is Pine will hang at sending the email,
> and while my session is waiting, I'll telnet in in another window. I'll
> connect, but I won't get the login prompt. About 30 seconds later, just as
> Pine wakes up and sends the message, I'll get presented with the login
> Not exactly a scientific explanation, but I've been poking around the
> system and I can't seem to find anything else amiss (such as mutltiple
> connections hanging in FIN_WAIT).
> Has anyone had experience with similar hangs in Solaris 2.5 with or
> without sendmail being involved? I'm running on a Sparc 10 Dual M51 with
> 64 megs of ram, recommended patches installed (as of August).
> Thanks for any insight anyone can offer.
I received responses that pointed to DNS as the blame. I'm not quite sure
that DNS is at fault. The symtoms I am experiencing point to almost a
complete hang of the network stack. As I said previously, if I go to send
a message, the process sending the message (Pine in this case) hangs, and
during that hang the machine will not accept any telnet connections to it.
I can test this out by trying to telnet to the machine. I will connect,
but will not get a login prompt. Then, just when Pine stops hanging, the
telnet sessions will immediately issue a login prompt.
Thanks to all who have replied.
Here are some of the responses I've received:
>From email@example.com Tue Mar 10 07:36:12 1998
Date: Wed, 04 Mar 1998 08:51:55 -0800
From: Charlie Mengler <firstname.lastname@example.org>
To: Admin Staff <email@example.com>
Subject: Re: Sendmail 8.8.7 Hangs 2.5.1
If the "hangs" are showing up on both email & telnet,
how can that be a problem with sendmail? The conclusion
about sendmail being part of the problem does not seem
too reasonable to me.
If I had this type of problem, I'd be looking at DNS
-- Charlie Mengler firstname.lastname@example.org Anchor Chips Network Manager 619-613-7916(V) 619-676-6896(F) 12396 World Trade Drive #212 San Diego, CA 92128-3788
>From email@example.com Tue Mar 10 07:36:18 1998 Date: Wed, 4 Mar 1998 09:04:32 -0800 (PST) From: Joel Lee <firstname.lastname@example.org> To: Admin Staff <email@example.com> Subject: Re: Sendmail 8.8.7 Hangs 2.5.1
I had this problem, but it was due largely in part to DNS timing out. You should start that first. Once sendmail works, then you would start point finger else where.
-- Joel firstname.lastname@example.org
>From email@example.com Tue Mar 10 07:36:22 1998 Date: Wed, 4 Mar 1998 12:16:12 -0500 (EST) From: "Snr SysAdmin, Random Access" <firstname.lastname@example.org> To: Admin Staff <email@example.com> Subject: Re: Sendmail 8.8.7 Hangs 2.5.1
I run approximately the same system as you, a Sparc 10 with dual M51's and 128Meg of RAM. The sendmail configs can be tailored very finely and in fact you can set some variables to tell sendmail not to send outbound if the load average of the computer is too high. That -sounds- like the problem you're having, perhaps sending a huge file outwards that slows the system down immensely.
I'd actually suggest getting sendmail 8.8.8 since it works better than 8.8.7, and is far more configurable than usual. If you are not using m4 configs I'd also suggest using them too, I just switched to them and it has made my life -much- easier ;)
Have fun and I hope your problem gets fixed.
R. Dows <firstname.lastname@example.org>
>From email@example.com Tue Mar 10 07:36:27 1998 Date: Wed, 4 Mar 1998 12:22:10 -0500 (EST) From: Scott Morris <firstname.lastname@example.org> To: Admin Staff <email@example.com> Subject: Re: Sendmail 8.8.7 Hangs 2.5.1
Check out your DNS. I have run into this behavior when the system was doing a lookup on the domain. Sendmail stalls and then suddenly takes off, regardless of version.
----- Scott Morris GTE Telecommunication Services Inc. Information Security firstname.lastname@example.org
>From email@example.com Tue Mar 10 07:36:31 1998 Date: Wed, 4 Mar 1998 10:25:40 -0700 (MST) From: "Jeffrey K. Pado" <firstname.lastname@example.org> To: Admin Staff <email@example.com> Subject: Re: Sendmail 8.8.7 Hangs 2.5.1
You may want to check out the following page:
Seems similar to your problem.
Jeffrey K. Pado System Administrator "You've got to be a little crazy CDI CAD Services to keep from going insane!!!!!" Phoenix, AZ -- Famous Detroit DJ firstname.lastname@example.org (602)431-8036 x317
>From email@example.com Tue Mar 10 07:36:35 1998 Date: Wed, 4 Mar 1998 12:59:23 -0500 From: "Boland, John" <firstname.lastname@example.org> To: 'Admin Staff' <email@example.com> Subject: RE: Sendmail 8.8.7 Hangs 2.5.1
we had a similar problem here on a 6000/14cpu box. at first, sendmail would stop sending for short periods about to about 30 minutes. we added another office to the system and sendmail practically stopped working ( 2-3 hour delivery times). turns out that sendmail stop sending when system load average runs high, uptime was showing 15+. there is a command line option (typically x and X) that will adjust the send and receive load values.
hope this helps!!!
>From Andy.Cranston@thomson-directories.co.uk Tue Mar 10 07:36:41 1998 Date: Thu, 5 Mar 1998 09:21:32 -0000 From: "Cranston, Andy" <Andy.Cranston@thomson-directories.co.uk> To: 'Admin Staff' <firstname.lastname@example.org> Subject: RE: Sendmail 8.8.7 Hangs 2.5.1
Does your Sparc have a 10/100 Mbps Ethernet card? We have several SPARCstation5 workstations on a dedicated 100 Mbps Ethernet switch. Trying to telnet them from a 16 Mbps Token Ring network (also switched) used to display similar symptoms. Either you would get a connection as normal or you would get a telnet connected but no login prompt. Waiting 30-40 seconds sometimes cured the probem. Another way was to login to the workstation itself (after a trip up a flight of stairs!) and then from the workstation ping something (anything) on the Token Ring network. Now a telnet was possible with no delay.
I fixed this my putting a line in /etc/inittab as follows:
pi:234:respawn:/usr/sbin/ping -I 5 tfarsql1 >/dev/null 2<>/dev/null </dev/null
The host tfarsql1 is a server on our 16 Mbps Token Ring network. This continual ping seems to "keep alive" the connection via the switch and we have not seen this problem again. Running ping every five seconds is hardly a system overhead and, although crude, does, as I say, do the trick. Something similar for you might work.
I can't put this behaviour down to Solaris and my gut feeling is that our LAN is not joining the 100 Mbps Ethernet and 16 Mbps Token Ring switches together correctly.
If you are running 100 Mbps Ethernet to slower network technologies it might be worth investigating the network infrastructure first.
Hopefully this might give you a little bit more food for thought.
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:12:33 CDT