stack trace page with kdump, possible?

2009-01-14 Thread Gianluca Cecchi
Hello,
with CentOS 4 I was able to setup a server 1 with netdump client
package and a netdump server 2 so that when a panic occured on 1, I
received on the netdump server 2 both the vmcore file and another file
named "log" containing stack trace (this log on 2 was also filled by
commands such as "echo m > /proc/sysrq-trigger" issued on 1).
With Fedora 10 and CentOS 5 I now have kdump. I configured it on both
kinds of OS so that I successfully get an scp of the vmcore when I
force a panic with
echo c > /proc/sysrq-trigger.
Unfortunately I don't get the log file with the stack trace page and
this would be useful for example if the crash happens during the night
and the display goes stbdy or if the server immediately reboots and
you don't have the time to see the stack trace page.
Any help if it is possible to configure kdump so that it repostrs also
the stack page?
Or if I can easily get the stack page from the vmcore file?

Thanks in advance,
Gianluca

-- 
fedora-list mailing list
fedora-list@redhat.com
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines


Re: stack trace page with kdump, possible?

2009-01-14 Thread Bryn M. Reeves

Gianluca Cecchi wrote:

Hello,
with CentOS 4 I was able to setup a server 1 with netdump client
package and a netdump server 2 so that when a panic occured on 1, I
received on the netdump server 2 both the vmcore file and another file
named "log" containing stack trace (this log on 2 was also filled by
commands such as "echo m > /proc/sysrq-trigger" issued on 1).
With Fedora 10 and CentOS 5 I now have kdump. I configured it on both
kinds of OS so that I successfully get an scp of the vmcore when I
force a panic with
echo c > /proc/sysrq-trigger.
Unfortunately I don't get the log file with the stack trace page and
this would be useful for example if the crash happens during the night
and the display goes stbdy or if the server immediately reboots and
you don't have the time to see the stack trace page.
Any help if it is possible to configure kdump so that it repostrs also
the stack page?
Or if I can easily get the stack page from the vmcore file?


Although network crash dumps via netdump have been removed from the 
distribution, the log messages can still be captured using the 
netconsole facility. See the configuration file in 
/etc/sysconfig/netconsole and the initscript /etc/init.d/netconsole.


You'll also need a syslog or netdump server configured on the network to 
capture the messages.


Regards,
Bryn.

--
fedora-list mailing list
fedora-list@redhat.com
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines


Re: stack trace page with kdump, possible?

2009-01-14 Thread Gianluca Cecchi
On Wed, 14 Jan 2009 11:35:12 + Bryn M. Reeve wrote:
> Although network crash dumps via netdump have been removed from the 
> distribution, the log messages can still be captured using the netconsole 
> facility.
> See the configuration file in /etc/sysconfig/netconsole and the initscript 
> /etc/init.d/netconsole.
> You'll also need a syslog or netdump server configured on the network to 
> capture the messages.
Ok, it works. thanks.
If the client is centos5/f10 and so does need netconsole service, the
remote should be syslogd server, netdump server alone is not
sufficient.
Now if I want the logs of the clients on different files and not all
in the same /var/log/messages of the syslog server, how can I get it?
By default I get lines with ip of the clients to understand the source
(.49 and .50) of the type

Jan 14 17:24:00 10.4.5.49 SysRq :
Jan 14 17:24:00 10.4.5.49 Show Memory
Jan 14 17:24:00 10.4.5.49 Mem-info:
Jan 14 17:24:00 10.4.5.49 Node 0
Jan 14 17:24:00 10.4.5.49 DMA per-cpu:
Jan 14 17:24:00 10.4.5.49
Jan 14 17:24:00 10.4.5.49 cpu 0 hot: high 0, batch 1 used:0
...
Jan 14 17:25:00 10.4.5.50 SysRq :
Jan 14 17:25:00 10.4.5.50 Show Memory
Jan 14 17:25:00 10.4.5.50 Mem-info:
Jan 14 17:25:00 10.4.5.50 Node 0
Jan 14 17:25:00 10.4.5.50 DMA per-cpu:
Jan 14 17:25:00 10.4.5.50
Jan 14 17:25:00 10.4.5.50 cpu 0 hot: high 0, batch 1 used:0

can I have different directories for different clients?
I did not find an option in syslog.conf for client or server...

Tahnks
Gianluca

-- 
fedora-list mailing list
fedora-list@redhat.com
To unsubscribe: https://www.redhat.com/mailman/listinfo/fedora-list
Guidelines: http://fedoraproject.org/wiki/Communicate/MailingListGuidelines