mnl@idtsun.e-technik.th-darmstadt.de (Michael N. Lipp) (06/20/91)
Hi, we have SunOS-4.1-sendmail running on two hosts, marlowe being the mailhost, spade a client with its spool/mail NFS-mounted. We repeatedly get "net hang"-messages like these: Jun 20 02:35:02 marlowe sendmail[9572]: AA09569: locked Jun 20 02:58:01 marlowe sendmail[9569]: AA09569: SYSERR: net hang reading from spade.dtro.e-technik.th-darmstadt.de: Connection timed out during collect with spade.dtro.e-technik.th-darmstadt.de Jun 20 02:58:02 marlowe sendmail[9569]: AA09569: message-id=<9106200028.AA09569@digtec.dtro.e-technik.th-darmstadt.de> Jun 20 02:58:02 marlowe sendmail[9569]: AA09569: from=<root>, size=22056, class=0 Can anybody tell me what "net hang" is? The two are on the same ethernet (and spade being /usr-NFS-client couldn`t run without marlowe anyway). So the net should work. When tested by hand, mailing on the machines works flawlessly. Looking at the dates, someone suggested that it is "batch -m" that sends the offending messages. But what could batch do when execing sendmail that makes sendmail fail that way? Thanks in advance Michael -- -----------------,------------------------------,------------------------------ Michael N. Lipp ! Institut fuer Datentechnik ! Phone: 49-6151-163776 ! Merckstr. 25 ,----------' Fax: 49-6151-164976 ! D-6100 Darmstadt ! E-Mail: (xdatmnlx@ddathd21.bitnet) ! (Germany) ! mnl@dtro.e-technik.th-darmstadt.de -----------------'-------------------'-----------------------------------------
rickert@mp.cs.niu.edu (Neil Rickert) (06/20/91)
In article <MNL.91Jun20142856@mnlsun.idtsun.e-technik.th-darmstadt.de> mnl@idtsun.e-technik.th-darmstadt.de (Michael N. Lipp) writes: >Jun 20 02:58:01 marlowe sendmail[9569]: AA09569: SYSERR: net hang reading from spade.dtro.e-technik.th-darmstadt.de: Connection timed out during collect with spade.dtro.e-technik.th-darmstadt.de > >Can anybody tell me what "net hang" is? The two are on the same This usually occurs when there is some miscommunication and both ends of the SMTP connection are reading and waiting for the other end to send something. >ethernet (and spade being /usr-NFS-client couldn`t run without marlowe >anyway). So the net should work. When tested by hand, mailing on the >machines works flawlessly. Looking at the dates, someone suggested >that it is "batch -m" that sends the offending messages. But what >could batch do when execing sendmail that makes sendmail fail that way? As a rough guess, based on its name, perhaps "batch" is sending a "batch" of SMTP commands. As defined, SMTP is a dialogue, with a response for each SMTP command. After sending a command, you are supposed to wait for the response before sending anything more. The design of sendmail depends on this behavior. Failure to observe it can lead to exactly this kind of net hang. -- =*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*= Neil W. Rickert, Computer Science <rickert@cs.niu.edu> Northern Illinois Univ. DeKalb, IL 60115 +1-815-753-6940