https://bugzilla.novell.com/show_bug.cgi?id=663367
https://bugzilla.novell.com/show_bug.cgi?id=663367#c2
Elmar Stellnberger changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|NEEDINFO |NEW
InfoProvider|estellnb@gmail.com |
--- Comment #2 from Elmar Stellnberger 2011-01-12 11:14:21 UTC ---
1. If there was a crash: Certainly there often is; for several reasons: kernel
oopses, crashes and hangs, Xorg event queue overflows & lifelocks, etc.
(This is simply because Linux software is often tested by the end user.)
2. If there was a crash I want to report it to get it resolved.
3. If I want to report a crash I will be in need of the logfiles.
4. The most important logfiles like /var/log/messages and /var/log/warn are
constantly destroyed as soon as you reboot into the system that just has
crashed. This is because these files are newly created (overwritingly) on
bootupup after kernel initialization (/var/log/boot.msg). They do and should
only contain information related to a single boot-session (i.e.
bootup-work-shutdown/crash).
5. A rescue system can help because it boots a system from CD without touching
your hard drive so that everything in your /var partition will still stay in
place. Simply mount it then and evacuate all you need on an USB-stick or to
your home or another place in your var-partition. That`s it!
Bad luck only if you don`t succeed to make your bios boot from CD instead of
hdd in time. Then all the information is lost - and many bugs are hard to
reproduce.
Consequently:
We need to create a backup of at least /var/log/messages and /var/log/warn
right after kernel initialization but before these files are re-created
overwriting their old ancestors.
Reasonsing:
The same happens with /var/log/boot.msg (logged during kernel initialization
at the very beginning). It is automatically back-uped to /var/log/boot.omsg.
Sameways /var/log/Xorg.0.log is backuped to /var/log/Xorg.0.log.old on X
startup. For the logfiles created by rsyslogd (warn, messages & others) we will
however have to initiate this by hand. RSysLogD simply does not do that
(automatically) on its own. As far as I have questioned this by various people
using other distros logrotate is assumed to do be conceived for that purpose.
What needs to be done:
The most challenging part of it will for sure be to call logrotate just
before messages & warn are re-created.
For that purpose it will be useful to configure logrotate appropriately (The
configuration file for the current (not very sensful) cron-based logrotate
invocation locates in /etc/logrotate.d/syslog); comparingly we can state:
* Extending by date is not useful here, as a machine will typically crash
multiple times per day (i.e. no dateext).
* no compress: We want to view the logfiles of the current day by an editor of
choice without having to unpack them. This is useful for a quick first hand
error outlook or analysis.
* I presmue that basically a rotate of 2 will suffice for most purposes. We
don`t have more than one backup for Xorg.log and boot.msg as well. However you
are free to let more being kept by default including additional packed
hyperlogbackups tagged by date.
reproduce by unplugging your power cable or just by pressing Alt-PrnScr-S-U-B.
P.S.: The text above got sorrystandingly not very entendible because I have
pressed commit by accident before being able to revise and correct it.
--
Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are on the CC list for the bug.