Topic: Too many connections: 40 >=40. Waiting one second.

[jim7jim]

Last week I started having problems with sending emails. Upon investigation I noticed messages similar to this in my qpsmtpd log:

2007-09-18 07:31:03.625297500 9442 Too many connections: 40 >= 40.  Waiting one second.
2007-09-18 07:31:04.628229500 9442 Too many connections: 40 >= 40.  Waiting one second.
2007-09-18 07:31:05.631127500 9442 Too many connections: 40 >= 40.  Waiting one second.
2007-09-18 07:31:06.633082500 9442 Too many connections: 40 >= 40.  Waiting one second.
2007-09-18 07:31:07.635041500 9442 Too many connections: 40 >= 40.  Waiting one second.
2007-09-18 07:31:08.637124500 9442 Too many connections: 40 >= 40.  Waiting one second.
2007-09-18 07:31:08.768775500 print() on closed filehandle GEN15547 at /usr/bin/qpsmtpd-forkserver line 291, <STDIN> line 2.
2007-09-18 07:31:08.769121500 14304 Connection Timed Out
2007-09-18 07:31:09.644647500 9442 cleaning up after 14304
2007-09-18 07:31:09.644651500 9442 Too many connections: 40 >= 40.  Waiting one second.


After searching the forum and bugtracker, I decided to update my SME Server. I was running v7.1. So last night I updated to version 7.2, restored the the default yum respositories and applied the post v7.2 updates.

I'm still getting the messages above in the qpsmtpd log. When this happens my users cannot send emails. I found from one of the forum posts that disabling the DNSBL might help. That is what I did last week and this week as a bandaid so that we can send emails.

I could never really find a definite solution in my searches. Does anyone know what my problem is?

TIA

[pfloor]

Check the bug tracker.  This bug http://bugs.contribs.org/show_bug.cgi?id=3352 has 101 replies and is probably most relevant.

Another thing to consider is the amount of memory your server has.  Spamassassin and clamav are very taxing on your system due to the amount of spam floating around these days.  If it is used even moderately it will need at least 512 Meg.  Any "business" type server should have at least 1 Gig of memory.

[jim7jim]

Thanks pfloor. Per your bug tracker link it looks like I need to apply a fix by updating from:

smeserver-qpsmtpd-1.2.1-43.el4.sme
to
smeserver-qpsmtpd-1.2.1-47.el4.sme

However, I can't seem to find the second package from: yum list updates

I tried looking in some repositories that I don't have enabled like smetest and smedev but I still can't find the smeserver-qpsmtpd-1.2.1-47.el4.sme package.

Does someone know where I can find it and the command I use to install?

TIA

Modified:
Oh by the way ... server resources don't appear to be an issue. This is a 3 Ghz Xeon w/ 1 GB RAM. RAM usage is only 250-300 MB and Load average per htop is usually < .3.

[byte]

Does someone know where I can find it and the command I use to install?

You can install via smeupdates-testing by typing...

yum upgrade --enable smeupdates-testing smeserver-qpsmtpd

then follow the usual steps in rebooting.

[jim7jim]

Thank you byte.

smeserver-qpsmtpd-1.2.1-47.el4.sme has been installed and so far it looks like it has fixed the problem.

Cheers! 

[pmiranda]

I have had the same problem for a couple of days, since last Monday evening.

I noticed that the qmstpd daemon was not fast enough to cope with all the connection requests it received. Incrementing the upper connection limit was pointless, whatever the value it was reached in minutes.

My SME Server runs on an HP DL 380 G4 bi-processor machine with 4 GB RAM, and the "top" command showed very low load average, low memory and no swapfile usage.

After many tests and tentatives, last Wednesday I installed smeserver-qpsmtpd-1.2.1-47.el4.sme from the smeupdates-testing repository, and the problem was solved: now the mail service is blazingly fast.

The mail services are configured as follows:

# config show qpsmtpd
qpsmtpd=service
    Bcc=disabled
    BccMode=cc
    BccUser=maillog
    DNSBL=enabled
    LogLevel=8
    MaxScannerSize=25000000
    RBLList=zen.spamhaus.org:whois.rfc-ignorant.org:dnsbl.njabl.org
    RHSBL=disabled
    RequireResolvableFromHost=yes
    SBLList=dsn.rfc-ignorant.org
    access=public
    status=enabled

# config show qmail
qmail=service
    MaxMessageSize=15000000
    status=enabled

If you have some comments about my current configuration, please feel free to express them.


Just a side note: I tried (erroneously, I know) to increase the concurrent qmail connections from 10 internal and 20 external to 20 internal and 200 external, using the qmail service configuration parameters ConcurrencyLocal and ConcurrencyRemote.

ConcurrencyLocal is not managed through the configuration file, the value 10 is hardcoded in the template file; so I changed its content.
ConcurrencyRemote is managed through the configuration file, instead.

After having signaled the email-update event, I checked the /var/qmail/control/concurrencylocal and /var/qmail/control/concurrencyremote values, and verified that both the values were the same I set above.

But looking at the qmail log it seems that these values are ignored: it keeps on saying

... status: local 1/10 remote 0/20
... status: local 2/10 remote 0/20
... status: local 1/10 remote 0/20
... status: local 2/10 remote 0/20
... status: local 1/10 remote 0/20
... status: local 2/10 remote 0/20
... status: local 1/10 remote 0/20
... status: local 0/10 remote 0/20

and so on, showing a connection limit of 10 and 20 whichever the values I set as said before.

Am I missing something, or did I hit a [known] bug?

[mmccarn]

I see the same behavior on my system.  I've opened a couple bugs:
qmail templates are unusual
signal-event email-update does not re-read qmail concurrency


If this issue is near and dear to your heart please add yourself to the CC list for the bugs.

[pmiranda]

If this issue is near and dear to your heart please add yourself to the CC list for the bugs.

I am not particularly fond of this bug  , but looking at my qmail logs I noticed that it often uses all the 10 available connection, and I was trying to increase them.

I would like that the e-smith template system honored the qmail configuration ConcurrencyLocal parameter - it is a very important aspect for a fully templated system as SME Server is - but thanks to the comments I read in the bug reports you mentioned, I managed to increase the qmail local limit anyway:

# mkdir -p /etc/e-smith/templates-custom/var/qmail/control
# echo 20 > /etc/e-smith/templates-custom/var/qmail/control/concurrencylocal
# signal-event email-update
# sv -v restart qmail

Now, my qmail log says that the available connections are 20, as expected:

...
... status: local 0/10 remote 0/20
... status: exiting
... status: local 0/20 remote 0/20
... status: local 1/20 remote 0/20
... status: local 2/20 remote 0/20
... status: local 1/20 remote 0/20
...

Thank you very much.

[CharlieBrady]

looking at my qmail logs I noticed that it often uses all the 10 available connection, and I was trying to increase them.

That may not be wise. If mail handling is slow, then increasing the number of concurrent connections may make it slower still.

The connection limit exists to protect your system from being overtaxed - if you just increase the connection limit, you remove that protection.

[pmiranda]

Thank you, I am aware of the issues involved in increasing the local connection limit.

Our internal mail traffic is high, way higher than the external one, and the SME Server in question is the only mail server in our net.

At the moment, the server seems always fast, responsive and with plenty of resources available; I am monitoring the qmail logs anyway: if the traffic hits the limit again, I will investigate further.