- Why not just log to syslogd?
-
Simply put, during the majority of the bootup process, syslogd
isn't running. Furthermore, starting it sooner won't catch some
messages, as syslogd requires writable / and /var/log directories.
- Why don't you just write directly to a log file, instead
of waiting for syslogd?
-
During the boot process, there could be multiple instances
of logging going on simultaneously. To correcty do this would
require implementing file locking in whatever was writing
the log files, and was deemed to be an unneccessary complication.
Additionally, by using syslogd to record all messages, this allows
for forwarding of messages to different log files, network hosts,
certain users, and other options.
- So what exactly is this [ OK ] / [ FAILED ] stuff,
anyways?
-
In the process of doing this logging, we are keeping track
of whether all the services/actions succeed or fail. It was
decided that service successes or failures should
be easily detectable by the user, as well as changing the
appearance of the bootup process from the "ungodly holy
mess of hackery that it is". Basically, everything that will
be logged as 'succeeded' gets an [ OK ], everything that
will be logged as 'failed' gets a [FAILED]. [PASSED] is
currently a special case used for successful filesystem
checks.
- That's nice. How do I turn it off?
-
If you edit /etc/sysconfig/init, you can change the
behavior of the bootup process. The BOOTUP variable
is the most useful one currently available. It supports the following
settings:
- color
-
the new, Red Hat 6.0 style bootup.
- verbose
-
the old, Red Hat 5.2 style bootup.
- serial (or currently, anything else)
-
the new style bootup, but with no
screen escapes for color or cursor positioning. Mostly useful
for serial consoles and the like.
More keywords may be defined as time goes on. Additionally,
through customizing the other variables present, you can change
(for example) what column to move to before printing a label,
or the colors/formatting used. For example, by editing the
'SETCOLOR_FAILURE' variable, it could be possible to make all
failures blink.
In any event, everything will still be logged.
- I get these strange messages from nfsd/apmd that screw
up the formatting. What's up with that?
-
These messages are messages from syslog and/or kernel
messages that are a high enough priority that they are echoed
to the screen. When the bootup starts, the initial console
loglevel is set to 1 (nothing but emergency messages) via
the loglevel program. However, once klogd and syslogd start,
this behavior is determined by their configuration. In general,
the simple solution is to patch the offending code if it bothers
you.
- Service foo hangs every time I boot - can I boot without it?
-
Yes; just type in linux confirm at the boot prompt, and
you will be asked for confirmation before starting services.
- Why aren't the messages in order?
-
This can't really be avoided. What happens at boot is:
- console-message/etc. -> buffered until syslogd starts
- kernel messages -> buffered until klogd starts
Then, when syslogd starts, buffer #1
is dumped to it; at the same time
klogd starts
and dumps buffer #2; meanwhile,
the bootup process continues, possibly
generating more messages.
Hence, the logs can end up somewhat jumbled.
Another minor issue is that the timestamps will change on the
messages if the time is reset during the install (via timezone
changes, or somesuch).
|
