"Diskless Nodes with Gentoo", September 16, 2005 version: Code Listing 6.4 shows typical slave runlevels. This list should be brought up-to-date, and things that are not relevant should be left out; they are just confusing in a troubleshooting situation. 1) The boot runlevel Remove "portmap" (it is not there in an up-to-date system) Remove "serial" (not relevant to simplest-possible diskless boot) Add "modules", "domainname", "checkroot" (they would typically be there) 2) The default runlevel Remove "clock" (it is not there on an up-to-date system) Remove "distccd" (not relevant to diskless boot) Remove "metalog" (not relevant to diskless boot) Remove "ntp-client" (not relevant to diskless boot) Remove "ntpd" (not relevant to diskless boot) Rename "vcron" to "vixie-cron" Add "syslog-ng" (it would typically be there)
portmap need not be mentioned as it has never been in a default runlevel afaik modules/checkroot are already run in boot, you dont need to add them metalog vs syslog-ng depends completely on the user and seeing as how metalog is the default syslogger in Gentoo, mentioning that instead of syslog-ng makes more sense
(In reply to comment #1) > and seeing as how metalog is the default syslogger in Gentoo, > mentioning that instead of syslog-ng makes more sense "The following example installs syslog-ng. Of course substitute with your system logger" (hb-install-next)
(In reply to comment #1) > metalog vs syslog-ng depends completely on the user and seeing as how metalog is > the default syslogger in Gentoo, mentioning that instead of syslog-ng makes more > sense I see now that the Handbook mentions three loggers (sysklogd, metalog, syslog-ng) and it is obviously up to the user to choose one. But where is it said that metalog is the default syslogger in Gentoo? The Gentoo install documents currently promotes syslog-ng over the others to some little extent: the Handbook provides a code listing for syslog-ng, and the Quick Install Guide mentions syslog-ng but none of the others.
run `emerge virtual/logger -p` on an empty system ... it will give back metalog
(In reply to comment #4) > run `emerge virtual/logger -p` on an empty system ... it will give back metalog OK then, metalog it is. Thanks for pointing me to this (little documented?) feature .. I realize now that the file /usr/portage/profiles/base/virtuals would have saved me some guesswork in the past. Actually it could be used even in the Diskless Nodes document: "emerge virtual/tftp" would be more generic than "emerge tftp-hpa".
FYI, ad syslog-ng vs. metalog: http://thread.gmane.org/gmane.linux.gentoo.devel/31841
Created attachment 78535 [details, diff] diskless-howto.xml.patch Done as requested. (Also, there are lots of GuideXML coding style fixes to be applied as well, but this is just a content change.) :) Patch is attached.
Cool patch, but metalog is still the default logger. It seems that thread on -dev just died. Anyone care to bring it up again? I've already brought up the logrotate thing recently, I don't want to draw more flames on me!
(In reply to comment #8) > Cool patch, but metalog is still the default logger. It seems that thread on > -dev just died. Anyone care to bring it up again? I've already brought up the > logrotate thing recently, I don't want to draw more flames on me! > Is there any reason why we should mention a logger at all? We know that: 1. Right now the default logger (virtual/logger) is metalog. 2. Gentoo users have choice, and lots of 'em prefer syslog-ng. 3. This code listing only shows *typical* slave runlevels. Therefore, we can have the more generic setup (that seems to be preferred) by just not mentioning any logger at all. Take out the metalog/syslog-ng init.d line altogether. Would this work? I'm hoping to just add the rest of the fixes and get this bug closed without a huge discussion of the merits/problems of one logger over the other. :)
In CVS. Thanks for reporting.
