mail very slow

[Starrynight_1]

Hi all

I have a problem with my mail, that is it takes about 5 minutes to execute the comand "mail ???@???.?? < anyfile". So after some investigation I found that if I stop sendmail deamon, rm /dev/log file, kill any pedding sendmail processes and run the deamon again the problem was resolved. Which I did...but the next day the /dev/log was there again and the problem came back with him :-(.
Can anyone tell what /dev/log is all about and also what can I do to take care of this problem for good...

Thanks in advance.

SN

[Jordan Bean]

You deleted /dev/log?!?! This is a FIFO read by syslogd. You'd better restart syslodg imediately to recreate and reopen it. Here is a list of some daemons that do write to it:

named
sendmail
bootpd
rpcbind
nfsd
nfsd
nfsd
inetd
pwgrd
nfsd

[Jordan Bean]

Wait... If /dev/log exists and is a plain file open by the other daemons (fuser /dev/log will tell you), this it would be safest to recover this way:

1. Stop sendmail since each child will open /dev/log for writing... creating it if it does not already exist.

2. Rename /dev/log to /dev/log.old. This will allow the other daemons to continue because the file is already open and the inode hasn't changed.

3. Restart syslogd to recreate /dev/log as a FIFO.

4. Reboot to safely restart all the other daemons on the FIFO.

[Starrynight_1]

Ops!
Well, I asked an HP-UX guru if I could do it, and than deleted /dev/log.......
Thank you for your explanation but I really need help with the mail problem. What could be the problem? Do you have any idea?

SN

[Jordan Bean]

mail and mailx pipe messages to non-daemon invocations of sendmail and wait until they are finished. Let's correct the /dev/log problem first just in case each sendmail is waiting to write to /dev/log. If the delay persists, then step up LogLevel to 13 in /etc/mail/sendmail.cf, restart sendmail, and watch /var/adm/syslog/mail.log. (Expect a lot of debug info.) Hopefully we'll get some indication of the problem.

[Starrynight_1]

Hello all

Thanks to eveyone that answered my question. I sholud say that all of you helped me, after I do as you told the problem is resolved but after sometime cames back again.
I run fuser /dev/log and all the I get is the process ID of syslogd deamon.
Can you tell what is causing this situation?

Regards,
SN

[Fabrice Meynard]

Hi,
i don't know what can be the link between /dev/log and sendmail (mail) performance problem. I already had performance problems with this product due to processor load average.
By default, sendmail only queue messages when load average is over 8 and refuse any messages whan load average is over 12. You can change these value in /etc/mail/sendmail.cf
When the problem happened on your server, check processor load average.
You spoke about sendmail process. You only need this daemon to accept (receive) email, not to send email.

Hope this helps,
Fabrice

[Jordan Bean]

Sorry about the delay. How are things today? How's the load on your system?

The sendmail daemon will do two things: listen for SMTP on a configurable port (default 25) and process the queue at configurable time intervals. The former task can be disabled (don't use -bd), but you will need sendmail running at least to process the queue when messages are defered (eg use -q30m) due to excessive system load or connection failures.

Have you tried increasing the LogLevel?