Re: Event logging vs enhancing printk

Francois-Xavier Kowalski (francois-xavier_kowalski@hp.com)
Wed, 10 Apr 2002 19:13:34 +0200


Hello,

Michel Dagenais <michel.dagenais@polymtl.ca> writes

>> People want to be able to get better debug information, with more detail
>> than is currently possible with printk, hence the Event logging project.
>
>I would like to emphasize that point; logging and tracing is of prime
>importance in several areas:
>
>- Telecom and Security. Logging and auditing is a requirement in Telecom
> applications. It is for instance one of the important features of the
> Distributed Security Infrastructure proposed by Ericsson Research and to
> be presented in a couple of months at the Ottawa Linux Symposium.
> The OSDL "Carrier Grade Linux" is certainly no different.
>

I am currently working in the Telecom area (signalling), on systems that
usually process several tenths of thousands of telephone calls per
second, during years. So I am the kind of guy interrested in having
enhancements to the Linux logging sub-systems, to:

1. Make my life easier when something goes wrong on my user's sites &
I am supposed to gather the information from the logs.
2. Make my user's life easier, to detect that something is going
wrong on my machines (among hunderds of other ones), so that he
can (try to) join me.
3. Make my developer's life easier, not forcing me to code yet
another log parser dedicated to one system & use open standard
instead.
4. Not make my Linux desktop system an intensive disk & CPU consume r
for logging-only purpose (this is more a personal view than a
professional constraint... :-)

I have been following the work done by evlog team since a few monthes
now, so I can say now - as an evlog user - that it provides support for
every requirements listed above, the following ways:

* Evlog can be configured to flag log messages as part of a
functional area (facilty) so that they can be logged in separated
log files, regardless of the source of the log: kernel or
userland. Take the case of a telecom protocol stack running partly
in kernel space (device driver and/or protocol module) and partly
in userland (application). Evlog can be configured a way so that
logs coming from the telecom functional area are consolidated at
the same place. If something goes wrong in kernel space, the logs
of the whole telecom sub-system are placed together showing
(mostly) the appropriated order of failure. This considerably
helps troubleshooting a failure (my trouble-shooting life).
* From the telecom network supervision/management perspective (my
user's life), the ability of evlog to have registered call-backs
(in user-land) on specified (configured) events permits to:
o Take automatic corrective actions if possible
o Raise network alarms, so that the operators can log on the
system & trouble-shoot it (or at least apply a corrective
procedure & call a help-desk)
* Evlog is a proposed standard to POSIX. I do hope that it will be
accepted (as well as in the LSB, but this is probably another
story...), so that the telecom stack & network supervision work
can be re-used from one system to another.
* Evlog can be totally optimized away at kernel configuration time
(like any other CONFIG_XXX component, after all), or have no
noticeable no cost when compiled-in & powered-off. (my day-to-day
Linux desktop user life).

For the reasons listed above, I strongly support integrating evlog in
the Linux kernel.

FiX

-- 
Francois-Xavier "FiX" KOWALSKI     /_ __    Tel: +33 (0)4 76 14 63 27
Telecom Infrastructure Division   / //_/    Fax: +33 (0)4 76 14 43 23
SigTech eXpert                      /       HP Telnet: 779-6327
                               i n v e n t  http://www.hp.com/go/opencall

- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/