devil-linux-develop Mailing List for Devil-Linux (Page 23)

Heiko,

On 12/06/2009 05:51 AM, Heiko Zuerker wrote:
> Serge,

> The ipvsadm compile fails and as far as I can tell it’s because of the
> missing libnl package.
> 
> Did you already create a build script for it?
> 
> http://www.infradead.org/~tgr/libnl/
Oops. Sure, it was done. As it has been testing several months I completely
forgot about this dependency. I'll fix it in a couple of minutes.

I'm sorry,
Serge

Serge,

 

The ipvsadm compile fails and as far as I can tell it's because of the
missing libnl package.

Did you already create a build script for it?

http://www.infradead.org/~tgr/libnl/

 

Regards

  Heiko Zuerker

  http://www.devil-linux.org

 

Make[2]: Entering directory `/data/build/tmp/ipvsadm-1.25'

make[2]: warning: jobserver unavailable: using -j1.  Add `+' to parent make
rule.

make -C libipvs

make[3]: Entering directory `/data/build/tmp/ipvsadm-1.25/libipvs'

gcc -Wall -Wunused -Wstrict-prototypes -g -fPIC -DLIBIPVS_USE_NL
-DHAVE_NET_IP_VS_H -c -o libipvs.o libipvs.c

In file included from libipvs.h:13,

                 from libipvs.c:23:

ip_vs.h:15:29: error: netlink/netlink.h: No such file or directory

ip_vs.h:16:31: error: netlink/genl/genl.h: No such file or directory

ip_vs.h:17:31: error: netlink/genl/ctrl.h: No such file or directory

In file included from libipvs.h:13,

                 from libipvs.c:23:

ip_vs.h:510: error: array type has incomplete element type

ip_vs.h:511: error: array type has incomplete element type

ip_vs.h:512: error: array type has incomplete element type

ip_vs.h:513: error: array type has incomplete element type

ip_vs.h:514: error: array type has incomplete element type

ip_vs.h:515: error: array type has incomplete element type

On 12/04/2009 07:42 AM, Andrzej Odyniec wrote:

> 
> The main difference, I think, is Quagga version -- now 0.99.15 patched for 
> realms using diffs for 0.99.12.
> I will try now to patch it (mainly bgpd) by newest accessible set of Adrian 
> Ban patches based on Calin Velea patches for realms tested for 0.99.14.
Thanks for pointing to the problem - I've updated the patch in DL sources. Patch
for 0.99.14 is completely compatible  with 0.99.15, the difference is only the
line with version.


--
Serge

Hi,

This night I tried to change DL on my main router replacing 13-jul to 1-dec 
compilation. And after change my addresses in Warsaw was unreachable from 
south of Poland. After returning to 13-jul image, all functionality returned 
to previous, correct state and all tunnels from south of Poland (Katowice, 
Kraków) was restored.

The main difference, I think, is Quagga version -- now 0.99.15 patched for 
realms using diffs for 0.99.12.

I will try now to patch it (mainly bgpd) by newest accessible set of Adrian 
Ban patches based on Calin Velea patches for realms tested for 0.99.14.

But mayby anyone know another reason, why simple replacing 0.99.12 to 0.99.15 
changed BGPD broadcasting? Maybe this is configuration trick? I don't use 
realms settings.

Regards

-- 
Andrzej Odyniec

> -----Original Message-----
> From: Bruce Smith [mailto:bw...@re...]
> Sent: Friday, December 04, 2009 8:23 AM
> To: dev...@li...
> Subject: Re: [Devil-linux-develop] beat that build time!
> 
> > Yesterday I upgraded my CPU in my vmware server (ESXI 4.0u1)  to an
> AMD
> > Phenom II 3.2Ghz.
> 
> Just a single CPU machine?
> 
> I'm not familiar with AMD.  How many cores is that CPU?

It's a 4 core CPU.
http://www.newegg.com/Product/Product.aspx?Item=N82E16819103674
 
> > I have a VM with 3 CPUs and 1GB of memory, which I use to build DL.
> >
> > The nightly build of the grsec version finished (without the make
> prepare)
> > in 2 hours and 14 minutes!
> 
> How long did it take previously?
> What CPU did you have previously?

Before I had a AMD Phenom I with 2.5 GHz (also 4 core), which took about 3
1/2 hours.

Heiko

> Yesterday I upgraded my CPU in my vmware server (ESXI 4.0u1)  to an AMD
> Phenom II 3.2Ghz.

Just a single CPU machine?

I'm not familiar with AMD.  How many cores is that CPU?

> I have a VM with 3 CPUs and 1GB of memory, which I use to build DL.
>
> The nightly build of the grsec version finished (without the make prepare)
> in 2 hours and 14 minutes!

How long did it take previously?
What CPU did you have previously?

 - BS

Hey,

 

I just gotta share that with you guys..

 

Yesterday I upgraded my CPU in my vmware server (ESXI 4.0u1)  to an AMD
Phenom II 3.2Ghz.

I have a VM with 3 CPUs and 1GB of memory, which I use to build DL.

The nightly build of the grsec version finished (without the make prepare)
in 2 hours and 14 minutes!

 

Here are the settings I'm currently using:

CONFIG_PMAKE=2

CONFIG_PARALLEL_JOBS=3

 

Regards

  Heiko Zuerker

  http://www.devil-linux.org

 

> -----Original Message-----
> From: Heiko Zuerker [mailto:he...@zu...]
> Sent: Sunday, November 29, 2009 3:23 PM
> To: dev...@li...
> Subject: Re: [Devil-linux-develop] compiling new version
> 
> > -----Original Message-----
> > From: Heiko Zuerker [mailto:he...@zu...]
> > Sent: Sunday, November 29, 2009 3:13 PM
> > To: dev...@li...
> > Subject: Re: [Devil-linux-develop] syslog-ng version
> >
> > > -----Original Message-----
> > > From: Bruce Smith [mailto:bw...@re...]
> > > Sent: Sunday, November 29, 2009 3:05 PM
> > > To: dev...@li...
> > > Subject: Re: [Devil-linux-develop] syslog-ng version
> > >
> > > Have you committed all of your fixes, in case I want to start a
> > > compile?
> >
> > I have a new compile running with a couple of software updates
> > (including
> > the 2.6.31.6 kernel), but that'll run for a couple more hours before
> I
> > know
> > if it even works. I may not check that in until tomorrow.
> > So I'd say go ahead.
> 
> I just realized that spamassassin is broken in CVS.
> I have to wait for my current build to finish, before I can fix it.

The spamassassin problem should be fixed.
I won't upload any of the other changes until tomorrow, since I need to do
more testing.

Heiko

> -----Original Message-----
> From: Heiko Zuerker [mailto:he...@zu...]
> Sent: Sunday, November 29, 2009 3:13 PM
> To: dev...@li...
> Subject: Re: [Devil-linux-develop] syslog-ng version
> 
> > -----Original Message-----
> > From: Bruce Smith [mailto:bw...@re...]
> > Sent: Sunday, November 29, 2009 3:05 PM
> > To: dev...@li...
> > Subject: Re: [Devil-linux-develop] syslog-ng version
> >
> > Have you committed all of your fixes, in case I want to start a
> > compile?
> 
> I have a new compile running with a couple of software updates
> (including
> the 2.6.31.6 kernel), but that'll run for a couple more hours before I
> know
> if it even works. I may not check that in until tomorrow.
> So I'd say go ahead.

I just realized that spamassassin is broken in CVS.
I have to wait for my current build to finish, before I can fix it.

Heiko

I'll wait awhile and start it just before bedtime.
Please check in tonight if everything works.

 - BS


On Sun, Nov 29, 2009 at 16:12, Heiko Zuerker <he...@zu...> wrote:
>> -----Original Message-----
>> From: Bruce Smith [mailto:bw...@re...]
>> Sent: Sunday, November 29, 2009 3:05 PM
>> To: dev...@li...
>> Subject: Re: [Devil-linux-develop] syslog-ng version
>>
>> Have you committed all of your fixes, in case I want to start a
>> compile?
>
> I have a new compile running with a couple of software updates (including
> the 2.6.31.6 kernel), but that'll run for a couple more hours before I know
> if it even works. I may not check that in until tomorrow.
> So I'd say go ahead.
>
> Heiko

> -----Original Message-----
> From: Bruce Smith [mailto:bw...@re...]
> Sent: Sunday, November 29, 2009 3:05 PM
> To: dev...@li...
> Subject: Re: [Devil-linux-develop] syslog-ng version
> 
> Have you committed all of your fixes, in case I want to start a
> compile?

I have a new compile running with a couple of software updates (including
the 2.6.31.6 kernel), but that'll run for a couple more hours before I know
if it even works. I may not check that in until tomorrow.
So I'd say go ahead.

Heiko

Have you committed all of your fixes, in case I want to start a compile?

 - BS


> I think it's finally fixed.
> The /etc/init.d/jail script changed the syslog-ng.conf and then reloaded the
> config. This also explained why it only seemed to happen after upgrades and
> not after normal reboots.
> I took that section completely out, since all the relevant changes are
> supposed to be done by upgrade-config anyway.
>
> I'll uploaded a new version if my nightly build finishes successfully.
>
> Heiko

I think it's finally fixed.
The /etc/init.d/jail script changed the syslog-ng.conf and then reloaded the
config. This also explained why it only seemed to happen after upgrades and
not after normal reboots.
I took that section completely out, since all the relevant changes are
supposed to be done by upgrade-config anyway.

I'll uploaded a new version if my nightly build finishes successfully.

Heiko

* Heiko Zuerker (he...@zu...) wrote:
> > -----Original Message-----
> > From: Olivier Thauvin [mailto:nan...@na...]
> > Sent: Saturday, November 28, 2009 9:31 AM
> > To: dev...@li...
> > Subject: [Devil-linux-develop] Issue using perl module
> > 
> > Hi,
> > 
> > I do think some perl modules need to be rebuild.
> > 
> > This error occurs on Devil 1.4 rc2:
> > 
> > root@sphinx:~ # perl -MDBI -e "DBI->connect('dbi:Pg:')"
> > install_driver(Pg) failed: DBD::Pg object version undef does not match
> > bootstrap parameter 2.13.1 at /usr/lib/perl5/5.8.9/i686-linux-thread-
> > multi/DynaLoader.pm line 249.
> > Compilation failed in require at (eval 3) line 3.
> 
> Can you please create a bug report in Mantis (link on website), so we don't lose track of this?

Done:

https://sourceforge.net/apps/mantisbt/devil-linux/view.php?id=42

> 
> Thanks
>   Heiko
> 
> 
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Devil-linux-develop mailing list
> Dev...@li...
> https://lists.sourceforge.net/lists/listinfo/devil-linux-develop
-- 

Olivier Thauvin
CNRS  -  LATMOS
♖ ♘ ♗ ♕ ♔ ♗ ♘ ♖

> -----Original Message-----
> From: Olivier Thauvin [mailto:nan...@na...]
> Sent: Saturday, November 28, 2009 9:31 AM
> To: dev...@li...
> Subject: [Devil-linux-develop] Issue using perl module
> 
> Hi,
> 
> I do think some perl modules need to be rebuild.
> 
> This error occurs on Devil 1.4 rc2:
> 
> root@sphinx:~ # perl -MDBI -e "DBI->connect('dbi:Pg:')"
> install_driver(Pg) failed: DBD::Pg object version undef does not match
> bootstrap parameter 2.13.1 at /usr/lib/perl5/5.8.9/i686-linux-thread-
> multi/DynaLoader.pm line 249.
> Compilation failed in require at (eval 3) line 3.

Can you please create a bug report in Mantis (link on website), so we don't lose track of this?

Thanks
  Heiko

Hi,

I do think some perl modules need to be rebuild.

This error occurs on Devil 1.4 rc2:

root@sphinx:~ # perl -MDBI -e "DBI->connect('dbi:Pg:')"
install_driver(Pg) failed: DBD::Pg object version undef does not match
bootstrap parameter 2.13.1 at
/usr/lib/perl5/5.8.9/i686-linux-thread-multi/DynaLoader.pm line 249.
Compilation failed in require at (eval 3) line 3.

Regards.

-- 

Olivier Thauvin
CNRS  -  LATMOS
♖ ♘ ♗ ♕ ♔ ♗ ♘ ♖

I have syslog-ng reload messages in my log at the times of upgrades (and 
at no other times except when I did a reload manually). See below some 
extracts from the syslog-ng's log (discontinuities shown by '[snip]'):

Oct 25 10:19:43 src@Samba syslog-ng[2229]: syslog-ng shutting down; 
version='3.0.4' [this is shutdown]
Oct 25 11:22:19 src@Samba aufs 2-standalone.tree-31-20091012 [this is 
start of reboot with (I think) new version of DL 1.4RC2]
-- [snip] --
Oct 25 11:22:27 src@Samba xinetd[7703]: xinetd Version 2.3.14 started 
with no options compiled in.
Oct 25 11:22:27 src@Samba xinetd[7703]: Started working: 1 available service
Oct 25 10:22:34 src@Samba ntpdate[7712]: step time server 192.33.96.102 
offset -3601.043559 sec
Oct 25 10:22:35 src@Samba syslog-ng[7318]: Configuration reload request 
received, reloading configuration;
-- [snip] --
Oct 25 10:22:41 src@Samba ctnetlink v0.93: registering with nfnetlink.
Oct 25 10:22:47 src@Samba login[8865]: pam_unix(login:session): session 
opened for user root by LOGIN(uid=0)
Oct 25 10:22:47 src@Samba login[8896]: ROOT LOGIN on 'tty1'
Oct 25 10:23:34 src@Samba nmbd[7629]: [2009/10/25 10:23:34, 0] 
nmbd/nmbd.c:71(terminate)
Oct 25 10:23:34 src@Samba nmbd[7629]: Got SIGTERM: going down...
Oct 27 12:09:58 src@Samba syslog-ng[7318]: segfault at 10 ip b77237e5 sp 
bf866120 error 4 in syslog-ng[b76fa000+4f000]
-- [snip] --
Nov 9 19:04:53 src@Samba syslog-ng[2213]: syslog-ng shutting down; 
version='3.0.4' [this is shutdown]
Nov 9 19:08:37 src@Samba kernel: : aufs 2-standalone.tree-31-20091012 
[this is reboot with (I think) another DL 1.4RC2 upgrade]
-- [snip] --
Nov 9 19:08:46 src@Samba xinetd[7610]: xinetd Version 2.3.14 started 
with no options compiled in.
Nov 9 19:08:46 src@Samba xinetd[7610]: Started working: 1 available service
Nov 9 19:08:53 src@Samba ntpdate[7619]: step time server 91.194.67.1 
offset -0.999261 sec
Nov 9 19:08:54 src@Samba syslog-ng[7225]: Configuration reload request 
received, reloading configuration;
-- [snip] --
Nov 9 19:09:00 src@Samba kernel: : nf_conntrack.acct=1 kernel parameter, 
acct=1 nf_conntrack module option or
Nov 9 19:09:00 src@Samba kernel: : sysctl 
net.netfilter.nf_conntrack_acct=1 to enable it.
Nov 9 19:09:00 src@Samba kernel: : ctnetlink v0.93: registering with 
nfnetlink.
Nov 9 19:09:07 src@Samba login[8786]: pam_unix(login:session): session 
opened for user root by LOGIN(uid=0)
Nov 9 19:09:07 src@Samba login[8816]: ROOT LOGIN on 'tty1'
Nov 10 06:24:05 src@Samba kernel: : syslog-ng[7225]: segfault at 10 ip 
b772c8d5 sp bf95e070 error 4 in syslog-ng[b7703000+4f000]

I just noticed from these logs that the syslog-ng reloads are 
immediately preceded by ntpdate updates which push the time backwards 
(in the first case by over 1 hour and in the second case by 1 second). 
Could this be upsetting syslog-ng and forcing the reload? But syslog-ng 
appears to carry on working for a while, although it has stopped working 
long before it segfaults; in both cases there is a long gap between the 
preceding log entry and the segfault (and this should certainly not 
happen because lots of cron jobs should be running in the meantime and 
registering the log).

I have to admit I don't really know what chroot jails are, but I do have 
'Use the chroot jails' set to 'Yes' in DL setup, I do use postfix, and 
because I use samba at each upgrade I do have winbind running. However 
winbind fills the log with messages and doesn't seem to help me so I get 
rid of it (by substituting a modified /etc/init.d/samba) - so it only 
runs shortly after an upgrade and not thereafter.

Hope this gives someone a clue as to what is going on.

Dominic

Heiko Zuerker wrote:
>
> I looked through the script but couldn’t find anything which would 
> cause a reload.
>
> It may be a good idea if somebody else double checks.
>
> In the systems you experience the crashes, do you have anything 
> running in a chroot jail (bind, postfix,etc.) ?
>
> Heiko
>
> *From:* Dominic Raferd [mailto:dl...@ed...]
> *Sent:* Sunday, November 22, 2009 11:20 AM
> *To:* dev...@li...
> *Subject:* Re: [Devil-linux-develop] syslog-ng version
>
> Heiko Zuerker wrote:
>
>     I changed it back to dgram in CVS for now, so we can see if the next DL
>
>     upgrades behave better or if the problem lies somewhere else.
>
>         
>
>  
> Unix-stream was not the problem, I just upgraded a VM and syslog-ng died
> again.
>   
>
> Whatever the underlying cause I guess it is connected with the reload 
> command. Do we have any instances of the crash happening after 
> syslog-ng restart rather than reload? Maybe at upgrade the script 
> could just use syslog-ng restart instead of reload?
>
> Dominic
>
> ------------------------------------------------------------------------
>
> ------------------------------------------------------------------------------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008 30-Day 
> trial. Simplify your report design, integration and deployment - and focus on 
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> ------------------------------------------------------------------------
>
> _______________________________________________
> Devil-linux-develop mailing list
> Dev...@li...
> https://lists.sourceforge.net/lists/listinfo/devil-linux-develop
>   

I looked through the script but couldn't find anything which would cause a
reload.

It may be a good idea if somebody else double checks.

 

In the systems you experience the crashes, do you have anything running in a
chroot jail (bind, postfix,etc.) ?

 

Heiko

 

From: Dominic Raferd [mailto:dl...@ed...] 
Sent: Sunday, November 22, 2009 11:20 AM
To: dev...@li...
Subject: Re: [Devil-linux-develop] syslog-ng version

 

Heiko Zuerker wrote: 

I changed it back to dgram in CVS for now, so we can see if the next DL
upgrades behave better or if the problem lies somewhere else.
    

 
Unix-stream was not the problem, I just upgraded a VM and syslog-ng died
again.
  

Whatever the underlying cause I guess it is connected with the reload
command. Do we have any instances of the crash happening after syslog-ng
restart rather than reload? Maybe at upgrade the script could just use
syslog-ng restart instead of reload?

Dominic 

 

> -----Original Message-----
> From: Bruce Smith [mailto:bw...@re...]
> Sent: Sunday, November 22, 2009 11:33 AM
> To: dev...@li...
> Subject: Re: [Devil-linux-develop] syslog-ng version
> 
> > Whatever the underlying cause I guess it is connected with the reload
> > command. Do we have any instances of the crash happening after
> syslog-ng
> > restart rather than reload? Maybe at upgrade the script could just
> use
> > syslog-ng restart instead of reload?
> 
> Would it have anything to do with the order services are started upon
> boot?
> 
> Does something need to be started before syslog-ng?  Just a thought...

It only seems to happen after upgrades to a newer version, I don't think I
had it die after a normal reboot (need to verify that).

Heiko

> Whatever the underlying cause I guess it is connected with the reload
> command. Do we have any instances of the crash happening after syslog-ng
> restart rather than reload? Maybe at upgrade the script could just use
> syslog-ng restart instead of reload?

Would it have anything to do with the order services are started upon boot?

Does something need to be started before syslog-ng?  Just a thought...

 - BS

<!DOCTYPE html PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN">
<html>
<head>
  <meta content="text/html;charset=ISO-8859-1" http-equiv="Content-Type">
  <title></title>
</head>
<body bgcolor="#ffffff" text="#000000">
Heiko Zuerker wrote:
<blockquote cite="mid:000001ca6b7d$79a8f380$6cfada80$@org" type="cite">
  <blockquote type="cite">
    <pre wrap="">I changed it back to dgram in CVS for now, so we can see if the next DL
upgrades behave better or if the problem lies somewhere else.
    </pre>
  </blockquote>
  <pre wrap=""><!---->
Unix-stream was not the problem, I just upgraded a VM and syslog-ng died
again.
  </pre>
</blockquote>
Whatever the underlying cause I guess it is connected with the reload
command. Do we have any instances of the crash happening after
syslog-ng restart rather than reload? Maybe at upgrade the script could
just use syslog-ng restart instead of reload?<br>
<br>
Dominic
</body>
</html>

> I changed it back to dgram in CVS for now, so we can see if the next DL
> upgrades behave better or if the problem lies somewhere else.

Unix-stream was not the problem, I just upgraded a VM and syslog-ng died
again.

Heiko

> -----Original Message-----
> From: Serge Leschinsky [mailto:fi...@in...]
> Sent: Saturday, November 21, 2009 1:57 PM
> To: dev...@li...
> Subject: Re: [Devil-linux-develop] syslog-ng version
> 
> On 11/21/2009 04:26 AM, Heiko Zuerker wrote:
> > Hey,
> >
> > I just upgraded all my VMs to the latest and greatest DL and about
> half of
> > them had syslog-ng crash after not even a minute.
> > The syslog-ng doesn't even get reloaded during the upgrade, it's
> simply not
> > running when the files are changed.
> 
> :-(
> 
> > I'll refert back to unix-dgram and see what happens in the next
> upgrade.
> > Serge, do you remember why you switched it from unix-dgram to unix-
> stream?
> >
> Sure.
> There were 2 reasons - the first one is the explanation from syslog-ng
> 1.6
> documentation about messages loosing probability:
> 
> unix-stream() and unix-dgram()
> 
> These two drivers behave similarly: they open the given AF_UNIX socket
> and start
> listening on it for messages. unix-stream() is primarily used on Linux
> and uses
> SOCK_STREAM semantics (connection oriented, no messages are lost);
> while
> unix-dgram() is used on BSDs and uses SOCK_DGRAM semantics: this may
> result in
> lost local messages if the system is overloaded.
> 
> http://www.cs.utah.edu/~clake/docs/syslog-ng-1.6/#id2525653
> 
> and the second one is multiline messages handling ( I got the same
> result from
> my experiments (JBOSS sends multiline messages), but in the message
> below Bazsi
> explains the reason of the difference)
> 
> > In short, when multiline log messages sent via /dev/log socket they
> are
> > splitted into multiple log entries. Only first entry will have proper
> > timestamp and facility.
> >
> > This behavior differs from sysklogd and metalog, where multiple lines
> are
> > concatenated.
> > Even syslog-ng itself concatenates them when used via udp socket.
> >
> > Tested on syslog-ng-2.1.3.
> 
> I guess this depends on whether you are using unix-stream() or
> unix-dgram().
> 
> The first concatenates it, the 2nd doesn't. Simply because unix-dgram
> knows about the size of the datagrams whereas unix-stream doesn't.
> 
> https://lists.balabit.hu/pipermail/syslog-ng/2009-September/013330.html
> 
> 
> I switched to "unix-stream" because the both of them work pretty well
> and
> "stream" looks to be  a bit more attractive. If the problem is in this
> option
> let's revert it to "dgram".

Thanks for the explanation, that makes a lot of sense.

I changed it back to dgram in CVS for now, so we can see if the next DL
upgrades behave better or if the problem lies somewhere else.

Heiko

On 11/21/2009 04:26 AM, Heiko Zuerker wrote:
> Hey,
> 
> I just upgraded all my VMs to the latest and greatest DL and about half of
> them had syslog-ng crash after not even a minute.
> The syslog-ng doesn't even get reloaded during the upgrade, it's simply not
> running when the files are changed.

:-(

> I'll refert back to unix-dgram and see what happens in the next upgrade.
> Serge, do you remember why you switched it from unix-dgram to unix-stream?
> 
Sure.
There were 2 reasons - the first one is the explanation from syslog-ng 1.6
documentation about messages loosing probability:

unix-stream() and unix-dgram()

These two drivers behave similarly: they open the given AF_UNIX socket and start
listening on it for messages. unix-stream() is primarily used on Linux and uses
SOCK_STREAM semantics (connection oriented, no messages are lost); while
unix-dgram() is used on BSDs and uses SOCK_DGRAM semantics: this may result in
lost local messages if the system is overloaded.

http://www.cs.utah.edu/~clake/docs/syslog-ng-1.6/#id2525653

and the second one is multiline messages handling ( I got the same result from
my experiments (JBOSS sends multiline messages), but in the message below Bazsi
explains the reason of the difference)

> In short, when multiline log messages sent via /dev/log socket they are
> splitted into multiple log entries. Only first entry will have proper
> timestamp and facility.
>
> This behavior differs from sysklogd and metalog, where multiple lines are
> concatenated.
> Even syslog-ng itself concatenates them when used via udp socket.
>
> Tested on syslog-ng-2.1.3.

I guess this depends on whether you are using unix-stream() or
unix-dgram().

The first concatenates it, the 2nd doesn't. Simply because unix-dgram
knows about the size of the datagrams whereas unix-stream doesn't.

https://lists.balabit.hu/pipermail/syslog-ng/2009-September/013330.html


I switched to "unix-stream" because the both of them work pretty well and
"stream" looks to be  a bit more attractive. If the problem is in this option
let's revert it to "dgram".

Serge

Hey,

I just upgraded all my VMs to the latest and greatest DL and about half of
them had syslog-ng crash after not even a minute.
The syslog-ng doesn't even get reloaded during the upgrade, it's simply not
running when the files are changed.

I'll refert back to unix-dgram and see what happens in the next upgrade.
Serge, do you remember why you switched it from unix-dgram to unix-stream?

Heiko

> -----Original Message-----
> From: Dominic Raferd [mailto:dl...@ed...]
> Sent: Tuesday, November 17, 2009 5:11 AM
> To: dev...@li...
> Subject: Re: [Devil-linux-develop] syslog-ng version
> 
> Serge Leschinsky wrote:
> > On 11/15/2009 06:02 AM, Bruce Smith wrote:
> >
> >> A better question might be
> >> "Does syslog-ng run fine for anyone in the latest RC2?".
> >>
> >> Personally I'm not running syslog-ng on RC2.
> >> So asking me if it crashes does not apply to me, and you might
> >> incorrectly infer from my silence that it's working for me.
> >>
> >>  - BS
> >>
> >
> > My systems work correctly. However, these systems are not "official"
> DL build,
> > but I believe it's not a reason of the problem reproducibility...
> >
> > Serge
> >
> I now think that the problem may relate to syslog-ng switching upon
> reload between source unix-stream and source unix-dgram. The default
> conf has (stripping comments):
> 
> @version: 3.0
> options { long_hostnames(on); flush_lines(0); time_reopen(60); use_dns
> (no); };
> source src { file("/proc/kmsg" program_override("kernel: "));
> unix-stream("/dev/log" max-connections(1000)); internal(); };
> destination console { file("/dev/tty10"); };
> log { source(src); destination(console); };
> 
> My syslog-ng.conf has:
> 
> @version: 3.0
> options { stats_freq(7200); long_hostnames(on); flush_lines(0);
> time_reopen(60); use_dns (no);  };
> source src { file("/proc/kmsg" program_override("kernel: "));
> unix-dgram("/dev/log" max-connections(1000)); internal(); };
> destination console { file("/dev/tty10"); };
> log { source(src); destination(console); };
> destination logfile { file("/var/log/messages"); };
> log { source(src); destination(logfile); };
> 
> see some discussion here which may or may not be related:
> http://74.125.77.132/search?q=cache:j-kBNRHUY-EJ:www.mail-
> archive.com/debian-bugs-dist%40lists.debian.org/msg240161.html+unix-
> stream+unix-dgram+reload&cd=1&hl=en&ct=clnk&gl=uk
> 
> In any case my guess is that "/etc/init.d/syslog reload" after
> upgrading
> DL switches from unix-stream to unix-dgram (because it switches from
> the
> default configuration at boot time to my configuration which I
> naturally
> elect to retain), and that this causes some problem which, hours later,
> leads syslog-ng to crash. This would also explain why I cannot
> replicate
> the problem except at upgrade - and why others do not experience it.
> Also why Heiko found the problem disappeared after removing his custom
> syslog-ng.conf (if you still have the custom file somewhere, Heiko,
> maybe you could check if it had unix-dgram?)
> 
> I have changed my syslog-ng.conf to use unix-stream (the unix-dgram
> must
> be a carry-over from an earlier DL configuration), and am hoping that
> the problem will not recur.
> 
> BTW does anyone know what is /dev/tty10, is it working and how does one
> get to see it? I guess this is why I specified the additional
> destination logfile, because I had no idea where the console messages
> were going.
> 
> Dominic
> 
> -----------------------------------------------------------------------
> -------
> Let Crystal Reports handle the reporting - Free Crystal Reports 2008
> 30-Day
> trial. Simplify your report design, integration and deployment - and
> focus on
> what you do best, core application coding. Discover what's new with
> Crystal Reports now.  http://p.sf.net/sfu/bobj-july
> _______________________________________________
> Devil-linux-develop mailing list
> Dev...@li...
> https://lists.sourceforge.net/lists/listinfo/devil-linux-develop