Another Queue Question

Hi there, looking for some pointers,

Have a client that has had some serious downtime over the last few
days with their Linux Sendmail server.
The hardware has been fixed but we now have a situation that there are
280,000 files in /var/spool/mqueue.

The server is used for inbound email only and its directed to either
the clients Notes or MS Exchange infrastructure, all new email is
delivered quickly, just seems to be the old(er) messages that are
delayed, the number of files is decreasing but at a very slow rate.

The server is a Compaq DL380 with 6gb of memory, running linux with a
kernal of Linux 2.4.19-64GB-SMP.
The version of Sendmail is 3.0.6, this was a corporate image so I'm
not too sure on what open source build its based upon.

The current config used is very basic (so I am told):

divert(-1)dnl
# This file was created with Sendmail Switch.
divert(0)dnl
VERSIONID(`@(#)xxxx.m4 1.0 (sendmail) 1/30/2008')
MODIFY_MAILER_FLAGS(`LOCAL', `-S')dnl
define(`QUEUE_DIR', `/var/spool/mqueue/')dnl
define(`confBIND_OPTS', `+WorkAroundBrokenAAAA')dnl
define(`confLOG_LEVEL', `10')dnl
define(`confQUEUE_LA', `50')dnl
define(`confQUEUE_SORT_ORDER', `Modification')dnl
define(`confREFUSE_LA', `65')dnl
define(`confRUN_AS_USER', `smadmin')dnl
FEATURE(`access_db', `DATABASE_MAP_TYPE -T
MAIL_SETTINGS_DIR`'access')dnl
FEATURE(`blacklist_recipients')dnl
FEATURE(`generics_entire_domain')dnl
FEATURE(`genericstable', `DATABASE_MAP_TYPE
MAIL_SETTINGS_DIR`'genericstable')dnl
FEATURE(`mailertable', `DATABASE_MAP_TYPE
MAIL_SETTINGS_DIR`'mailertable')dnl
FEATURE(`no_default_msa')dnl
FEATURE(`nouucp', `reject')dnl
FEATURE(`relay_entire_domain')dnl
FEATURE(`use_cw_file')dnl
FEATURE(`virtusertable', `DATABASE_MAP_TYPE
MAIL_SETTINGS_DIR`'virtusertable')dnl
MAILER(`local')dnl
MAILER(`smtp')dnl

The Linux server in itself doesnt seem to be under too much load
either:

18:11:12 CPU %user %nice %system %idle intr/s
18:11:12 all 5.31 0.01 5.07 89.61 1233.16

Any pointers on what we can do to increase the throughput of this box
would be greatly appreciated..

Cheers

Al

Re: Another Queue Question

aldouk75-google@yahoo.co.uk writes:

> Hi there, looking for some pointers,
>
> Have a client that has had some serious downtime over the last few
> days with their Linux Sendmail server.
> The hardware has been fixed but we now have a situation that there are
> 280,000 files in /var/spool/mqueue.
>
> The server is used for inbound email only and its directed to either
> the clients Notes or MS Exchange infrastructure, all new email is
> delivered quickly, just seems to be the old(er) messages that are
> delayed, the number of files is decreasing but at a very slow rate.
>
> The server is a Compaq DL380 with 6gb of memory, running linux with a
> kernal of Linux 2.4.19-64GB-SMP.
> The version of Sendmail is 3.0.6, this was a corporate image so I'm
> not too sure on what open source build its based upon.
>
> The current config used is very basic (so I am told):
>
> divert(-1)dnl
> # This file was created with Sendmail Switch.
> divert(0)dnl
> VERSIONID(`@(#)xxxx.m4 1.0 (sendmail) 1/30/2008')
> MODIFY_MAILER_FLAGS(`LOCAL', `-S')dnl
> define(`QUEUE_DIR', `/var/spool/mqueue/')dnl
> define(`confBIND_OPTS', `+WorkAroundBrokenAAAA')dnl
> define(`confLOG_LEVEL', `10')dnl
> define(`confQUEUE_LA', `50')dnl
> define(`confQUEUE_SORT_ORDER', `Modification')dnl
> define(`confREFUSE_LA', `65')dnl
> define(`confRUN_AS_USER', `smadmin')dnl
> FEATURE(`access_db', `DATABASE_MAP_TYPE -T
> MAIL_SETTINGS_DIR`'access')dnl
> FEATURE(`blacklist_recipients')dnl
> FEATURE(`generics_entire_domain')dnl
> FEATURE(`genericstable', `DATABASE_MAP_TYPE
> MAIL_SETTINGS_DIR`'genericstable')dnl
> FEATURE(`mailertable', `DATABASE_MAP_TYPE
> MAIL_SETTINGS_DIR`'mailertable')dnl
> FEATURE(`no_default_msa')dnl
> FEATURE(`nouucp', `reject')dnl
> FEATURE(`relay_entire_domain')dnl
> FEATURE(`use_cw_file')dnl
> FEATURE(`virtusertable', `DATABASE_MAP_TYPE
> MAIL_SETTINGS_DIR`'virtusertable')dnl
> MAILER(`local')dnl
> MAILER(`smtp')dnl
>
> The Linux server in itself doesnt seem to be under too much load
> either:
>
> 18:11:12 CPU %user %nice %system %idle intr/s
> 18:11:12 all 5.31 0.01 5.07 89.61 1233.16
>
> Any pointers on what we can do to increase the throughput of this box
> would be greatly appreciated..

[ for "internet to internal" gateway]

First: Flush the queue faster using command below:
/usr/sbin/sendmail -q -OQueueSortOrder=random
More elaborate options required if you use more than four internal
destinations for SMTP connections.
Start *first* command with additional "-v" optional.
If queue is processed without porblems than start a few more (3-5)
(without -v) as background jobs.

Second:
1) setup separate queue group per every destination of
"to internal" SMTP connections with multiple queue directories in
every group (e.g. 4 directories)
2) configure sendmail to accept and process ETRN from intenal servers
It should allow the servers to request flushing queue upon then own
request without waiting for next queue run


--
[pl>en: Andrew] Andrzej Adam Filip : anfi@priv.onet.pl : anfi@xl.wp.pl
Open-Sendmail: http://open-sendmail.sourceforge.net/
Try to remove the color-problem by restarting your computer several times.
-- Microsoft Internet Explorer's README.TXT
----
http://groups.google.com/groups?selm=871w7zrwg2@minnie.fsf.h obby-site.com

Re: Another Queue Question

Andrzej Adam Filip wrote:
> aldouk75-google@yahoo.co.uk writes:
>
> > Hi there, looking for some pointers,
> >
> > Have a client that has had some serious downtime over the last few
> > days with their Linux Sendmail server.
> > The hardware has been fixed but we now have a situation that there are
> > 280,000 files in /var/spool/mqueue.
> >
> > The server is used for inbound email only and its directed to either
> > the clients Notes or MS Exchange infrastructure, all new email is
> > delivered quickly, just seems to be the old(er) messages that are
> > delayed, the number of files is decreasing but at a very slow rate.
> >
> > The server is a Compaq DL380 with 6gb of memory, running linux with a
> > kernal of Linux 2.4.19-64GB-SMP.
> > The version of Sendmail is 3.0.6, this was a corporate image so I'm
> > not too sure on what open source build its based upon.
> >
> > The current config used is very basic (so I am told):
> >
> > divert(-1)dnl
> > # This file was created with Sendmail Switch.
> > divert(0)dnl
> > VERSIONID(`@(#)xxxx.m4 1.0 (sendmail) 1/30/2008')
> > MODIFY_MAILER_FLAGS(`LOCAL', `-S')dnl
> > define(`QUEUE_DIR', `/var/spool/mqueue/')dnl
> > define(`confBIND_OPTS', `+WorkAroundBrokenAAAA')dnl
> > define(`confLOG_LEVEL', `10')dnl
> > define(`confQUEUE_LA', `50')dnl
> > define(`confQUEUE_SORT_ORDER', `Modification')dnl
> > define(`confREFUSE_LA', `65')dnl
> > define(`confRUN_AS_USER', `smadmin')dnl
> > FEATURE(`access_db', `DATABASE_MAP_TYPE -T
> > MAIL_SETTINGS_DIR`'access')dnl
> > FEATURE(`blacklist_recipients')dnl
> > FEATURE(`generics_entire_domain')dnl
> > FEATURE(`genericstable', `DATABASE_MAP_TYPE
> > MAIL_SETTINGS_DIR`'genericstable')dnl
> > FEATURE(`mailertable', `DATABASE_MAP_TYPE
> > MAIL_SETTINGS_DIR`'mailertable')dnl
> > FEATURE(`no_default_msa')dnl
> > FEATURE(`nouucp', `reject')dnl
> > FEATURE(`relay_entire_domain')dnl
> > FEATURE(`use_cw_file')dnl
> > FEATURE(`virtusertable', `DATABASE_MAP_TYPE
> > MAIL_SETTINGS_DIR`'virtusertable')dnl
> > MAILER(`local')dnl
> > MAILER(`smtp')dnl
> >
> > The Linux server in itself doesnt seem to be under too much load
> > either:
> >
> > 18:11:12 CPU %user %nice %system %idle intr/s
> > 18:11:12 all 5.31 0.01 5.07 89.61 1233.16
> >
> > Any pointers on what we can do to increase the throughput of this box
> > would be greatly appreciated..
>
> [ for "internet to internal" gateway]
>
> First: Flush the queue faster using command below:
> /usr/sbin/sendmail -q -OQueueSortOrder=random
> More elaborate options required if you use more than four internal
> destinations for SMTP connections.
> Start *first* command with additional "-v" optional.
> If queue is processed without porblems than start a few more (3-5)
> (without -v) as background jobs.
>
> Second:
> 1) setup separate queue group per every destination of
> "to internal" SMTP connections with multiple queue directories in
> every group (e.g. 4 directories)
> 2) configure sendmail to accept and process ETRN from intenal servers
> It should allow the servers to request flushing queue upon then own
> request without waiting for next queue run
>
>
> --
> [pl>en: Andrew] Andrzej Adam Filip : anfi@priv.onet.pl : anfi@xl.wp.pl
> Open-Sendmail: http://open-sendmail.sourceforge.net/
> Try to remove the color-problem by restarting your computer several times.
> -- Microsoft Internet Explorer's README.TXT
> ----
> http://groups.google.com/groups?selm=871w7zrwg2@minnie.fsf.h obby-site.com

Hi Andrew,
I came across a similiar problem back when sendmail8.11 was
involved. The workaround that we were advised to do was
* shutdown sendmail
* move out the files from mqueue
* restart sendmail but turn off acceptance of new email from the
outside
* move back the files to mqueue 4000 at a time
* turn back on acceptance of new email from the outside when last
batch back in
This was done on a 700mhz pc running RH9 with 768M RAM. Took less than
an hour.

Re: Another Queue Question

seaPea writes:

> Andrzej Adam Filip wrote:
>> aldouk75-google@yahoo.co.uk writes:
>>
>> > Hi there, looking for some pointers,
>> >
>> > Have a client that has had some serious downtime over the last few
>> > days with their Linux Sendmail server.
>> > The hardware has been fixed but we now have a situation that there are
>> > 280,000 files in /var/spool/mqueue.
>> >
>> > The server is used for inbound email only and its directed to either
>> > the clients Notes or MS Exchange infrastructure, all new email is
>> > delivered quickly, just seems to be the old(er) messages that are
>> > delayed, the number of files is decreasing but at a very slow rate.
>> >
>> > The server is a Compaq DL380 with 6gb of memory, running linux with a
>> > kernal of Linux 2.4.19-64GB-SMP.
>> > The version of Sendmail is 3.0.6, this was a corporate image so I'm
>> > not too sure on what open source build its based upon.
>> >
>> > The current config used is very basic (so I am told):
>> >
>> > divert(-1)dnl
>> > # This file was created with Sendmail Switch.
>> > divert(0)dnl
>> > VERSIONID(`@(#)xxxx.m4 1.0 (sendmail) 1/30/2008')
>> > MODIFY_MAILER_FLAGS(`LOCAL', `-S')dnl
>> > define(`QUEUE_DIR', `/var/spool/mqueue/')dnl
>> > define(`confBIND_OPTS', `+WorkAroundBrokenAAAA')dnl
>> > define(`confLOG_LEVEL', `10')dnl
>> > define(`confQUEUE_LA', `50')dnl
>> > define(`confQUEUE_SORT_ORDER', `Modification')dnl
>> > define(`confREFUSE_LA', `65')dnl
>> > define(`confRUN_AS_USER', `smadmin')dnl
>> > FEATURE(`access_db', `DATABASE_MAP_TYPE -T
>> > MAIL_SETTINGS_DIR`'access')dnl
>> > FEATURE(`blacklist_recipients')dnl
>> > FEATURE(`generics_entire_domain')dnl
>> > FEATURE(`genericstable', `DATABASE_MAP_TYPE
>> > MAIL_SETTINGS_DIR`'genericstable')dnl
>> > FEATURE(`mailertable', `DATABASE_MAP_TYPE
>> > MAIL_SETTINGS_DIR`'mailertable')dnl
>> > FEATURE(`no_default_msa')dnl
>> > FEATURE(`nouucp', `reject')dnl
>> > FEATURE(`relay_entire_domain')dnl
>> > FEATURE(`use_cw_file')dnl
>> > FEATURE(`virtusertable', `DATABASE_MAP_TYPE
>> > MAIL_SETTINGS_DIR`'virtusertable')dnl
>> > MAILER(`local')dnl
>> > MAILER(`smtp')dnl
>> >
>> > The Linux server in itself doesnt seem to be under too much load
>> > either:
>> >
>> > 18:11:12 CPU %user %nice %system %idle intr/s
>> > 18:11:12 all 5.31 0.01 5.07 89.61 1233.16
>> >
>> > Any pointers on what we can do to increase the throughput of this box
>> > would be greatly appreciated..
>>
>> [ for "internet to internal" gateway]
>>
>> First: Flush the queue faster using command below:
>> /usr/sbin/sendmail -q -OQueueSortOrder=random
>> More elaborate options required if you use more than four internal
>> destinations for SMTP connections.
>> Start *first* command with additional "-v" optional.
>> If queue is processed without porblems than start a few more (3-5)
>> (without -v) as background jobs.
>>
>> Second:
>> 1) setup separate queue group per every destination of
>> "to internal" SMTP connections with multiple queue directories in
>> every group (e.g. 4 directories)
>> 2) configure sendmail to accept and process ETRN from intenal servers
>> It should allow the servers to request flushing queue upon then own
>> request without waiting for next queue run
>
> Hi Andrew,
> I came across a similiar problem back when sendmail8.11 was
> involved. The workaround that we were advised to do was
> * shutdown sendmail
> * move out the files from mqueue
> * restart sendmail but turn off acceptance of new email from the
> outside
> * move back the files to mqueue 4000 at a time
> * turn back on acceptance of new email from the outside when last
> batch back in
> This was done on a 700mhz pc running RH9 with 768M RAM. Took less than
> an hour.

Perl motto: There is more than one way to do things right :-)
My motto: Usually *WHY* is more important than *WHAT* :-)

With such number of messages in queue there are (AFAIR) two main source
of possible problems:
a) memory overload by data required to process the queue in "sorted" way
"QueueSortOrder=random" I suggested avoids this trap
b) file system performance problems with *MANY* files in a (queue) directory
[ file systems: e.g. ext2, ufs, ...]
*many* file system types do linear search for files in directory

I have addressed "a", recipe you have mentioned may address "b".

--
[pl>en: Andrew] Andrzej Adam Filip : anfi@priv.onet.pl : anfi@xl.wp.pl
Open-Sendmail: http://open-sendmail.sourceforge.net/
"Would I turn on the gas if my pal Mugsy were in there?"
"You might, rabbit, you might!"
-- Looney Tunes, Bugs and Thugs (1954, Friz Freleng)
----
http://groups.google.com/groups?selm=87myqnqb3i@james.fsf.ho bby-site.com

Re: Another Queue Question

On Jan 30, 9:04=A0pm, Andrzej Adam Filip wrote:
> seaPea writes:
> > Andrzej Adam Filip wrote:
> >> aldouk75-goo...@yahoo.co.uk writes:
>
> >> > Hi there, looking for some pointers,
>
> >> > Have a client that has had some serious downtime over the last few
> >> > days with their Linux Sendmail server.
> >> > The hardware has been fixed but we now have a situation that there ar=
e
> >> > 280,000 files in /var/spool/mqueue.
>
> >> > The server is used for inbound email only and its directed to either
> >> > the clients Notes or MS Exchange infrastructure, all new email is
> >> > delivered quickly, just seems to be the old(er) messages that are
> >> > delayed, the number of files is decreasing but at a very slow rate.
>
> >> > The server is a Compaq DL380 with 6gb of memory, running linux with a=

> >> > kernal of Linux 2.4.19-64GB-SMP.
> >> > The version of Sendmail is 3.0.6, this was a corporate image so I'm
> >> > not too sure on what open source build its based upon.
>
> >> > The current config used is very basic (so I am told):
>
> >> > divert(-1)dnl
> >> > # This file was created with Sendmail Switch.
> >> > divert(0)dnl
> >> > VERSIONID(`@(#)xxxx.m4 =A0 =A0 =A0 1.0 (sendmail) 1/30/2008')
> >> > MODIFY_MAILER_FLAGS(`LOCAL', `-S')dnl
> >> > define(`QUEUE_DIR', `/var/spool/mqueue/')dnl
> >> > define(`confBIND_OPTS', `+WorkAroundBrokenAAAA')dnl
> >> > define(`confLOG_LEVEL', `10')dnl
> >> > define(`confQUEUE_LA', `50')dnl
> >> > define(`confQUEUE_SORT_ORDER', `Modification')dnl
> >> > define(`confREFUSE_LA', `65')dnl
> >> > define(`confRUN_AS_USER', `smadmin')dnl
> >> > FEATURE(`access_db', `DATABASE_MAP_TYPE -T
> >> > MAIL_SETTINGS_DIR`'access')dnl
> >> > FEATURE(`blacklist_recipients')dnl
> >> > FEATURE(`generics_entire_domain')dnl
> >> > FEATURE(`genericstable', `DATABASE_MAP_TYPE
> >> > MAIL_SETTINGS_DIR`'genericstable')dnl
> >> > FEATURE(`mailertable', `DATABASE_MAP_TYPE
> >> > MAIL_SETTINGS_DIR`'mailertable')dnl
> >> > FEATURE(`no_default_msa')dnl
> >> > FEATURE(`nouucp', `reject')dnl
> >> > FEATURE(`relay_entire_domain')dnl
> >> > FEATURE(`use_cw_file')dnl
> >> > FEATURE(`virtusertable', `DATABASE_MAP_TYPE
> >> > MAIL_SETTINGS_DIR`'virtusertable')dnl
> >> > MAILER(`local')dnl
> >> > MAILER(`smtp')dnl
>
> >> > The Linux server in itself doesnt seem to be under too much load
> >> > either:
>
> >> > 18:11:12 =A0 =A0 CPU =A0 %user =A0 %nice %system =A0 %idle =A0 =A0int=
r/s
> >> > 18:11:12 =A0 =A0 all =A0 =A0 =A0 5.31 =A0 =A0 =A0 0.01 =A0 =A05.07 =
=A0 =A0 =A0 =A0 89.61 =A0 1233.16
>
> >> > Any pointers on what we can do to increase the throughput of this box=

> >> > would be greatly appreciated..
>
> >> [ for "internet to internal" gateway]
>
> >> First: Flush the queue faster using command below:
> >> =A0 /usr/sbin/sendmail -q -OQueueSortOrder=3Drandom
> >> More elaborate options required if you use more than four internal
> >> destinations for SMTP connections.
> >> Start *first* command with additional "-v" optional.
> >> If queue is processed without porblems than start a few more (3-5)
> >> (without -v) as background jobs.
>
> >> Second:
> >> 1) setup separate queue group per every destination of
> >> =A0 =A0"to internal" SMTP connections with multiple queue directories i=
n
> >> =A0 =A0every group (e.g. 4 directories)
> >> 2) configure sendmail to accept and process ETRN from intenal servers
> >> =A0 =A0It should allow the servers to request flushing queue upon then =
own
> >> =A0 =A0request without waiting for next queue run
>
> > Hi Andrew,
> > I came across a similiar problem back when sendmail8.11 was
> > involved. The workaround that we were advised to do was
> > =A0 =A0* shutdown sendmail
> > =A0 =A0* move out the files from mqueue
> > =A0 =A0* restart sendmail but turn off acceptance of new email from the
> > =A0 =A0 =A0outside
> > =A0 =A0* move back the files to mqueue 4000 at a time
> > =A0 =A0* turn back on acceptance of new email from the outside when last=

> > =A0 =A0 =A0batch back in
> > This was done on a 700mhz pc running RH9 with 768M RAM. Took less than
> > an hour.
>
> Perl motto: There is more than one way to do things right :-)
> My motto: Usually *WHY* is more important than *WHAT* :-)
>
> With such number of messages in queue there are (AFAIR) two main source
> of possible problems:
> a) memory overload by data required to process the queue in "sorted" way
> =A0 =A0"QueueSortOrder=3Drandom" I suggested avoids this trap
> b) file system performance problems with *MANY* files in a (queue) directo=
ry
> =A0 =A0[ file systems: e.g. ext2, ufs, ...]
> =A0 =A0*many* file system types do linear search for files in directory
>
> I have addressed "a", recipe you have mentioned may address "b".
>
> --
> [pl>en: Andrew] Andrzej Adam Filip : a...@priv.onet.pl : a...@xl.wp.pl
> Open-Sendmail:http://open-sendmail.sourceforge.net/
> "Would I turn on the gas if my pal Mugsy were in there?"
> "You might, rabbit, you might!"
> =A0 -- Looney Tunes, Bugs and Thugs (1954, Friz Freleng)
> ----
> http://groups.google.com/groups?selm=3D87myqnq...@james.fsf. hobby-site.com=
- Hide quoted text -
>
> - Show quoted text -

Thanks lads, your tips have been really helpful, appreciated..

Al

Re: Another Queue Question

aldouk75-google@yahoo.co.uk writes:

> On Jan 30, 9:04 pm, Andrzej Adam Filip wrote:
>> seaPea writes:
>> > Andrzej Adam Filip wrote:
>> >> aldouk75-goo...@yahoo.co.uk writes:
>>
>> >> > Hi there, looking for some pointers,
>>
>> >> > Have a client that has had some serious downtime over the last few
>> >> > days with their Linux Sendmail server.
>> >> > The hardware has been fixed but we now have a situation that there are
>> >> > 280,000 files in /var/spool/mqueue.
>>
>> >> > The server is used for inbound email only and its directed to either
>> >> > the clients Notes or MS Exchange infrastructure, all new email is
>> >> > delivered quickly, just seems to be the old(er) messages that are
>> >> > delayed, the number of files is decreasing but at a very slow rate.
>>
>> >> > The server is a Compaq DL380 with 6gb of memory, running linux with a
>> >> > kernal of Linux 2.4.19-64GB-SMP.
>> >> > The version of Sendmail is 3.0.6, this was a corporate image so I'm
>> >> > not too sure on what open source build its based upon.
>>
>> >> > The current config used is very basic (so I am told):
>>
>> >> > divert(-1)dnl
>> >> > # This file was created with Sendmail Switch.
>> >> > divert(0)dnl
>> >> > VERSIONID(`@(#)xxxx.m4       1.0 (sendmail) 1/30/2008')
>> >> > MODIFY_MAILER_FLAGS(`LOCAL', `-S')dnl
>> >> > define(`QUEUE_DIR', `/var/spool/mqueue/')dnl
>> >> > define(`confBIND_OPTS', `+WorkAroundBrokenAAAA')dnl
>> >> > define(`confLOG_LEVEL', `10')dnl
>> >> > define(`confQUEUE_LA', `50')dnl
>> >> > define(`confQUEUE_SORT_ORDER', `Modification')dnl
>> >> > define(`confREFUSE_LA', `65')dnl
>> >> > define(`confRUN_AS_USER', `smadmin')dnl
>> >> > FEATURE(`access_db', `DATABASE_MAP_TYPE -T
>> >> > MAIL_SETTINGS_DIR`'access')dnl
>> >> > FEATURE(`blacklist_recipients')dnl
>> >> > FEATURE(`generics_entire_domain')dnl
>> >> > FEATURE(`genericstable', `DATABASE_MAP_TYPE
>> >> > MAIL_SETTINGS_DIR`'genericstable')dnl
>> >> > FEATURE(`mailertable', `DATABASE_MAP_TYPE
>> >> > MAIL_SETTINGS_DIR`'mailertable')dnl
>> >> > FEATURE(`no_default_msa')dnl
>> >> > FEATURE(`nouucp', `reject')dnl
>> >> > FEATURE(`relay_entire_domain')dnl
>> >> > FEATURE(`use_cw_file')dnl
>> >> > FEATURE(`virtusertable', `DATABASE_MAP_TYPE
>> >> > MAIL_SETTINGS_DIR`'virtusertable')dnl
>> >> > MAILER(`local')dnl
>> >> > MAILER(`smtp')dnl
>>
>> >> > The Linux server in itself doesnt seem to be under too much load
>> >> > either:
>>
>> >> > 18:11:12     CPU   %user   %nice %system   %idle    intr/s
>> >> > 18:11:12     all       5.31       0.01    5.07         89.61   1233.16
>>
>> >> > Any pointers on what we can do to increase the throughput of this box
>> >> > would be greatly appreciated..
>>
>> >> [ for "internet to internal" gateway]
>>
>> >> First: Flush the queue faster using command below:
>> >>   /usr/sbin/sendmail -q -OQueueSortOrder=random
>> >> More elaborate options required if you use more than four internal
>> >> destinations for SMTP connections.
>> >> Start *first* command with additional "-v" optional.
>> >> If queue is processed without porblems than start a few more (3-5)
>> >> (without -v) as background jobs.
>>
>> >> Second:
>> >> 1) setup separate queue group per every destination of
>> >>    "to internal" SMTP connections with multiple queue directories in
>> >>    every group (e.g. 4 directories)
>> >> 2) configure sendmail to accept and process ETRN from intenal servers
>> >>    It should allow the servers to request flushing queue upon then own
>> >>    request without waiting for next queue run
>>
>> > Hi Andrew,
>> > I came across a similiar problem back when sendmail8.11 was
>> > involved. The workaround that we were advised to do was
>> >    * shutdown sendmail
>> >    * move out the files from mqueue
>> >    * restart sendmail but turn off acceptance of new email from the
>> >      outside
>> >    * move back the files to mqueue 4000 at a time
>> >    * turn back on acceptance of new email from the outside when last
>> >      batch back in
>> > This was done on a 700mhz pc running RH9 with 768M RAM. Took less than
>> > an hour.
>>
>> Perl motto: There is more than one way to do things right :-)
>> My motto: Usually *WHY* is more important than *WHAT* :-)
>>
>> With such number of messages in queue there are (AFAIR) two main source
>> of possible problems:
>> a) memory overload by data required to process the queue in "sorted" way
>>    "QueueSortOrder=random" I suggested avoids this trap
>> b) file system performance problems with *MANY* files in a (queue) directory
>>    [ file systems: e.g. ext2, ufs, ...]
>>    *many* file system types do linear search for files in directory
>>
>> I have addressed "a", recipe you have mentioned may address "b".
>
> Thanks lads, your tips have been really helpful, appreciated..
>
> Al

Could you post "for future readers reference" which advises have you
tried and what kind of results have you achieved?

--
[pl>en: Andrew] Andrzej Adam Filip : anfi@priv.onet.pl : anfi@xl.wp.pl
Open-Sendmail: http://open-sendmail.sourceforge.net/
Experience, n.:
Something you don't get until just after you need it.
-- Olivier
----
http://groups.google.com/groups?selm=87hcgvottk@ryan.fsf.hob by-site.com

Re: Another Queue Question

>With such number of messages in queue there are (AFAIR) two main source of possible problems:
>a) memory overload by data required to process the queue in "sorted" way
> "QueueSortOrder=random" I suggested avoids this trap
>b) file system performance problems with *MANY* files in a (queue) directory
> [ file systems: e.g. ext2, ufs, ...]
> *many* file system types do linear search for files in directory

Similarly to 'b', rename the queue directory and recreate a new queue directory with the same (old) namd, permissions, group and owner. This will shrink the size f the directory (# of inodes / blocks) on disk, making the new directory faster to parse. This is also a reason fore multiple queue directories.



Grant. . . .