Mailing List Archive

Looking for bughunters...
Hi all,

we are still working to hunt down the unfortunate and extremely
hard-to-find (to say the least...) segfault issue that some experience.
Peter Vrabec has created a great new RPM with debugging support. We
would appreciate if those of you experiencing the problem could give it
a try.

Full information can be found on the rsyslog site at

http://www.rsyslog.com/Article139.phtml

Your help is deeply appreciated!

Thanks,
Rainer

PS: as a reminder, I do more in-depth reports of what is happening in my
blog at http://rgerhards.blogspot.com - you may be interested to follow
it. I do this so that I do not pollute the mailing list with
too-in-depth information. Thanks.
Looking for bughunters... [ In reply to ]
Am Dienstag, den 27.11.2007, 10:54 +0100 schrieb Rainer Gerhards:
[...]
> Your help is deeply appreciated!
Short: won't work out of the box on RHEL5.1.

Long:
Download of RPM to RHEL5.1, installed libmudflap and libmudflabp-devel,
build and installed rsyslog, did not install the mysql-package.
Felt annoyed when I realised that the package moved /etc/rsyslog.conf
to /etc/rsyslog.conf.rpmnew and copied a new configuration
to /etc/rsyslog.conf. Changed back to my old configuration.

Did "/sbin/rsyslogd -n -o -r514 -m 0". The daemon starts and that's it.
Killed it, did "/opt/rsyslog/sbin/rsyslogd -n -o -r514 -m 0" (the
"original 1.19.10") and it logged fine. So both the command-line and the
config seem to be valid. But the modified rsyslog+mudflap just sits
there in a loop, waiting for something to happen (according to strace).


open("/etc/hosts", O_RDONLY) = 8
fcntl(8, F_GETFD) = 0
fcntl(8, F_SETFD, FD_CLOEXEC) = 0
fstat(8, {st_mode=S_IFREG|0644, st_size=234, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0)
= 0x2aaaaaaad000
read(8, "# Do not remove the following li"..., 4096) = 234
read(8, "", 4096) = 0
close(8) = 0
munmap(0x2aaaaaaad000, 4096) = 0
open("/etc/ld.so.cache", O_RDONLY) = 8
fstat(8, {st_mode=S_IFREG|0644, st_size=135025, ...}) = 0
mmap(NULL, 135025, PROT_READ, MAP_PRIVATE, 8, 0) = 0x2aaaaae7e000
close(8) = 0
open("/lib64/libnss_dns.so.2", O_RDONLY) = 8
read(8, "\177ELF\2\1\1\0\0\0\0\0\0\0\0\0\3\0>\0\1\0\0\0\240\17\0\0\0\0\0
\0"..., 832) = 832
futex(0x2aaaaacea840, FUTEX_WAIT, 2, NULL) = -1 EINTR (Interrupted
system call)
--- SIGWINCH (Window changed) @ 0 (0) ---
futex(0x2aaaaacea840, FUTEX_WAIT, 2, NULL) = -1 EINTR (Interrupted
system call)
--- SIGWINCH (Window changed) @ 0 (0) ---
futex(0x2aaaaacea840, FUTEX_WAIT, 2, NULL) = -1 EINTR (Interrupted
system call)
--- SIGWINCH (Window changed) @ 0 (0) ---
futex(0x2aaaaacea840, FUTEX_WAIT, 2, NULL


Last lines indicate some DNS-problem and in fact adding "-x" to the
commandline helps. Now I'm waiting for the next crash to send my
output^.

--
CU,
Patrick.
Looking for bughunters... [ In reply to ]
Am Dienstag, den 27.11.2007, 16:31 +0100 schrieb Patrick von der Hagen:
[...]
> Last lines indicate some DNS-problem and in fact adding "-x" to the
> commandline helps. Now I'm waiting for the next crash to send my
> output^.
No crash so far. The load is noticably higher, but still acceptable.
Now I killed the debugging-release and used precisely the same
commandline to start the regular release, which caused some crashes
here.
As I mentioned the debugging-release had some problems with DNS, so I
can't run it the same way I usually try to run rsyslog. So I'm not quite
sure wheter or not the results can actually be compared.
Anyway, I'll try the new command-line with 1.19.10 first, if that
crashes I'll switch to 1.19.11.

--
CU,
Patrick.
Looking for bughunters... [ In reply to ]
Patrick,

excellent, many thanks. We stay tuned for updates. I am very hopeful for
the compiler switch issue...

Rainer

> -----Original Message-----
> From: rsyslog-bounces at lists.adiscon.com [mailto:rsyslog-
> bounces at lists.adiscon.com] On Behalf Of Patrick von der Hagen
> Sent: Thursday, November 29, 2007 12:19 PM
> To: rsyslog-users
> Subject: Re: [rsyslog] Looking for bughunters...
>
> Am Dienstag, den 27.11.2007, 16:31 +0100 schrieb Patrick von der
Hagen:
> [...]
> > Last lines indicate some DNS-problem and in fact adding "-x" to the
> > commandline helps. Now I'm waiting for the next crash to send my
> > output^.
> No crash so far. The load is noticably higher, but still acceptable.
> Now I killed the debugging-release and used precisely the same
> commandline to start the regular release, which caused some crashes
> here.
> As I mentioned the debugging-release had some problems with DNS, so I
> can't run it the same way I usually try to run rsyslog. So I'm not
> quite
> sure wheter or not the results can actually be compared.
> Anyway, I'll try the new command-line with 1.19.10 first, if that
> crashes I'll switch to 1.19.11.
>
> --
> CU,
> Patrick.
> _______________________________________________
> rsyslog mailing list
> http://lists.adiscon.net/mailman/listinfo/rsyslog
Looking for bughunters... [ In reply to ]
Hi all,

just a small update: 1.19.10-debug worked for two days without problems.
1.19.10 crashed after 5 hours.
Now I'll give 1.19.11 a try....

--
CU,
Patrick.
Looking for bughunters... [ In reply to ]
Patrick,

ok - even though this sounds somewhat silly, I think this is good news.
It is evidence that the problem is not related to the option. That, too,
points into the direction of a threading issue, so it is consistent with
former findings.

What I now hope is that the rest of the analysis is true, too. In that
case, 1.19.11 should work...

Thanks for all your help. Much appreciated.
Rainer

> -----Original Message-----
> From: rsyslog-bounces at lists.adiscon.com [mailto:rsyslog-
> bounces at lists.adiscon.com] On Behalf Of Patrick von der Hagen
> Sent: Friday, November 30, 2007 11:24 AM
> To: rsyslog-users
> Subject: Re: [rsyslog] Looking for bughunters...
>
> Hi all,
>
> just a small update: 1.19.10-debug worked for two days without
> problems.
> 1.19.10 crashed after 5 hours.
> Now I'll give 1.19.11 a try....
>
> --
> CU,
> Patrick.
> _______________________________________________
> rsyslog mailing list
> http://lists.adiscon.net/mailman/listinfo/rsyslog