Outgoing Mail Stuck

[odie]

Hello

We are attempting to evaluate Scalix after previously evaluating Samsung Contact. With Samsung Contact, we were able to use our preferred MTA, Postfix. I have been able to configure Postfix and Scalix for incoming mail, but not outgoing. Outgoing mail simply sits in the UNIX queue until the unix.out daemon aborts after about a minute. The Scalix fatal error log reports:

SERIOUS ERROR Internet Mail (Outgoing ) 08.05.04 10:36:27
[OM 10270] Process about to terminate due to error.
Signal (Segmentation Violation) trapped by process 12374
Procedure trace follows:
<- mpl_AddMsgToQueue
<- Q_PutMsgOnQueue2
-> aud_AbortTransaction
<- aud_AbortTransaction
-> Q_PutMsgOnQueue2
-> Msg_GetTfHandle
<- Msg_GetTfHandle
-> tf_FlushFile
<- tf_FlushFile
-> Msg_GetRef
<- Msg_GetRef
-> mpl_AddMsgToQueue
-> gen_GetLock
<- gen_GetLock
-> mpl_OpenPoolFile
<- mpl_OpenPoolFile
Pid of logging process: 12374

And there are corresponding log entries in the mail log from postfix:

Aug 5 14:34:46 server postfix/smtpd[3386]: warning: connect #1 to subsystem private/proxymap: Permission denied
Aug 5 14:34:56 server postfix/smtpd[3386]: warning: connect #2 to subsystem private/proxymap: Permission denied
Aug 5 14:35:06 server postfix/smtpd[3386]: warning: connect #3 to subsystem private/proxymap: Permission denied
Aug 5 14:35:16 server postfix/smtpd[3386]: warning: connect #4 to subsystem private/proxymap: Permission denied
Aug 5 14:35:26 server postfix/smtpd[3386]: warning: connect #5 to subsystem private/proxymap: Permission denied
Aug 5 14:35:36 server postfix/smtpd[3386]: warning: connect #6 to subsystem private/proxymap: Permission denied
Aug 5 14:35:46 server postfix/smtpd[3386]: warning: connect #7 to subsystem private/proxymap: Permission denied
Aug 5 14:35:56 server postfix/smtpd[3386]: warning: connect #8 to subsystem private/proxymap: Permission denied
Aug 5 14:36:06 server postfix/smtpd[3386]: warning: connect #9 to subsystem private/proxymap: Permission denied
Aug 5 14:36:16 server postfix/smtpd[3386]: warning: connect #10 to subsystem private/proxymap: Permission denied
Aug 5 14:36:26 server postfix/smtpd[3386]: fatal: connect #11 to subsystem private/proxymap: Permission denied

I finally decided to try to use sendmail as my MTA to see if things improved. Actually, they got worse. :x As soon as the unix.out process starts, the service router queue begins to fill up with about 10-20 error messages per second. Then the local delivery queue begins to fill up with the messages from the service router. I end up having to kill the unix.out process. The Scalix fatal log reported something about being unable to access a Uri or Url database. Unfortunately those entries have already been overwritten. I do have what sendmail writes in the mail log file:

Aug 5 16:32:16 server sendmail[17775]: i75KWDXa017775: root@localhost did not issue MAIL/EXPN/VRFY/ETRN during connection to stdin

Any help would be greatly appreciated. I am at my wits end! :shock:

odie
[/quote]

[ScalixSupport]

Hi Odie,

First off, we don't support Postfix. It's on our list of considerations but there are no firm plans when this will happen.

Second, when you configured sendmail, did you make sure that the daemon was listening on 127.0.0.1 port 25 ? This is important for any outgoing mail whether it's from Scalix or from the Linux environment.

I'd be interested in the data from the fatal file. The logs get overwritten because they are cyclic but the fatal file is just a flat file so nothing should be overwritten.

Cheers

Dave Kelly
Senior Support Engineer
Scalix Corporation

[odie]

You're correct, of course. I found the log entries produced when I was running sendmail. They are:

WARNING Internet Mail (Outgoing ) 08.05.04 16:52:43
[OM 6001] Url Database is closed
-> tf_UnreadFile
<- tf_UnreadFile
-> tf_ReadRecord
-> tf_GetINT32
<- tf_GetINT32
<- tf_ReadRecord 1100 40
-> tf_GetINT32
<- tf_GetINT32
-> tf_GetINT16
<- tf_GetINT16
-> tf_GetINT32
<- tf_GetINT32
-> tf_GetINT32
<- tf_GetINT32
-> ul_ReadENU
<- /build/9.0.0.26-1/src/lib/ombase/ul/ul_ENU.c:56[3,6001]

I will check to make sure sendmail is configured per your specifications. As a side thought, if Scalix expects to talk to sendmail at 127.0.0.1 on port 25 using the SMTP protocol, why wouldn't it be able to talk to any other MTA, such as Postfix?

Odie

[odie]

I checked to see what process was listening at 127.0.0.1 on port 25. Postfix was listening, so I turned it off and enabled sendmail. The same thing happened. I then turned off sendmail so that nothing was listening on that port. The same thing happened.

I thought that the unix.out daemon doesn't connect to an internet socket, but uses whatever program is specified in the unix.mapper config file. In my case it is /usr/bin/sendmail -bs

Is this the problem? If I run /usr/bin/sendmail -bs I get the same output that I get when I telnet into sendmail.

Any help would be appreciated.

Odie

[ScalixSupport]

I did some checking. We shouldn't be seeing this error unless something went wrong prior to this.

The Url database that is being referred to is the USERLIST directory. If the database isn't open, it might indicate a permissions problem. What permissions does /opt/scalix/bin/unix.out have ?

WRT needing sendmail listening on 127.0.0.1 port 25, this isn't a Scalix requirement, it's a sendmail requirement. Since sendmail 8.12, the model for message submission from the local machine changed. Any command like mail or sendmail -bs submit their messages to 127.0.0.1:25 so the sendmail daemon must be listening on that port for any messages to be sent out of the machine.

Cheers

Dave

[odie]

The permissions on /opt/scalix/bin/unix.out are scalix:scalix and -r-xr-xr-x or 555. I configured sendmail to listen on the smtp port and the submission port. I can run /usr/bin/sendmail -bs from a command line and send mail to my own email address without a problem. However, as soon as I turn the Internet Mail Gateway back on, the queues start filling up with error messages again, and the original messages in the unix.out queue don't go anywhere.

Thanks for your help so far.

Odie

[odie]

I probably should have done this before, but I opened one of the 128 error messages I received this time. You are probably right about it being a permissions problem, but I don't know where to look. Here is what is says:

----------------------------- ERROR REPORT -----------------------------
Message could not be delivered to the following recipient:

[myusername] / internet
DDT1=RFC-822; DDV1=[my@emailaddress.com];

because: 'A mail transport failure occurred' [OM 9205]

Supplementary Information:

WARNING: RunAsGid for MSP ignored, check group ids (egid=0, want=51)


------------------------------------------------------------------------

Odie

[ScalixSupport]

This is definitely a permissions problem.

Take a look at the file /usr/share/doc/sendmail/SECURITY. There is a paragraph towards the top of the file that lists the expected permissions:


If you are running Sendmail 8.12.x, make sure you are using the following permissions as defined in sendmail/SECURITY:

-r-xr-sr-x root smmsp ... /PATH/TO/sendmail
drwxrwx--- smmsp smmsp ... /var/spool/clientmqueue
drwx------ root wheel ... /var/spool/mqueue
-r--r--r-- root wheel ... /etc/mail/sendmail.cf
-r--r--r-- root wheel ... /etc/mail/submit.cf

Cheers

Dave.

[odie]

Thanks a lot for your help, Dave! Resetting the permissions fixed the problem. Just for the record, I can't wait for Scalix to support Postfix.

Odie

[ScalixSupport]

Hello

We are attempting to evaluate Scalix after previously evaluating Samsung Contact. With Samsung Contact, we were able to use our preferred MTA, Postfix. I have been able to configure Postfix and Scalix for incoming mail, but not outgoing. Outgoing mail simply sits in the UNIX queue until the unix.out daemon aborts after about a minute. The Scalix fatal error log reports:

SERIOUS ERROR Internet Mail (Outgoing ) 08.05.04 10:36:27
[OM 10270] Process about to terminate due to error.
Signal (Segmentation Violation) trapped by process 12374
Procedure trace follows:
<- mpl_AddMsgToQueue
<- Q_PutMsgOnQueue2
-> aud_AbortTransaction
<- aud_AbortTransaction
-> Q_PutMsgOnQueue2
-> Msg_GetTfHandle
<- Msg_GetTfHandle
-> tf_FlushFile
<- tf_FlushFile
-> Msg_GetRef
<- Msg_GetRef
-> mpl_AddMsgToQueue
-> gen_GetLock
<- gen_GetLock
-> mpl_OpenPoolFile
<- mpl_OpenPoolFile
Pid of logging process: 12374

And there are corresponding log entries in the mail log from postfix:

Aug 5 14:34:46 server postfix/smtpd[3386]: warning: connect #1 to subsystem private/proxymap: Permission denied
Aug 5 14:34:56 server postfix/smtpd[3386]: warning: connect #2 to subsystem private/proxymap: Permission denied
Aug 5 14:35:06 server postfix/smtpd[3386]: warning: connect #3 to subsystem private/proxymap: Permission denied
Aug 5 14:35:16 server postfix/smtpd[3386]: warning: connect #4 to subsystem private/proxymap: Permission denied
Aug 5 14:35:26 server postfix/smtpd[3386]: warning: connect #5 to subsystem private/proxymap: Permission denied
Aug 5 14:35:36 server postfix/smtpd[3386]: warning: connect #6 to subsystem private/proxymap: Permission denied
Aug 5 14:35:46 server postfix/smtpd[3386]: warning: connect #7 to subsystem private/proxymap: Permission denied
Aug 5 14:35:56 server postfix/smtpd[3386]: warning: connect #8 to subsystem private/proxymap: Permission denied
Aug 5 14:36:06 server postfix/smtpd[3386]: warning: connect #9 to subsystem private/proxymap: Permission denied
Aug 5 14:36:16 server postfix/smtpd[3386]: warning: connect #10 to subsystem private/proxymap: Permission denied
Aug 5 14:36:26 server postfix/smtpd[3386]: fatal: connect #11 to subsystem private/proxymap: Permission denied

I finally decided to try to use sendmail as my MTA to see if things improved. Actually, they got worse. :x As soon as the unix.out process starts, the service router queue begins to fill up with about 10-20 error messages per second. Then the local delivery queue begins to fill up with the messages from the service router. I end up having to kill the unix.out process. The Scalix fatal log reported something about being unable to access a Uri or Url database. Unfortunately those entries have already been overwritten. I do have what sendmail writes in the mail log file:

Aug 5 16:32:16 server sendmail[17775]: i75KWDXa017775: root@localhost did not issue MAIL/EXPN/VRFY/ETRN during connection to stdin

Any help would be greatly appreciated. I am at my wits end! :shock:

odie

[/quote]

The solution to the proxymap problem is to change

#local_recipient_maps = proxy:unix:passwd.byname $alias_maps

to

local_recipient_maps =

This prevents Postfix from checking against passwd. It only works if you use Postfix as a MTA with no local users.

That said, Postfix is not officially supported. The reason why I post it here is that a customer reported the issue and everybody trying postfix will run into this.

Cheers,

Sascha.