readahead slows down boot
pere at hungry.com
Sat Apr 5 07:33:51 UTC 2008
> Petter Reinholdtsen wrote:
>>Perhaps it failed to stop when it should, and load all the files used when
>>you log in addition to the files that is used during boot? To test this
>>theory, you could run one profile run, reboot, and then try to boot
>>normally, without logging in after the profile run.
I found a machine to test on, and sure enough, the readahead-watch
process was still running after boot. So stop-readahead has failed to
kill it. I suspect the reason is because the pid file is missing or
contain the wrong pid. If this is the case, I would recommend
changing the pid file location used by readahead-watch
/var/run/readahead-watch.pid to /lib/init/rw/readahead-watch.pid.
Does it help to do 'echo RAMRUN=yes >> /etc/default/rcS'? It did when
I tested it.
> Profiling without logging in leads to the same results. Here is my
> /etc/boot generated by readahead:
Hm, about 1500 files to load during boot do not seem to wrong. Got
around 1200 on my laptop. Looking through the files, I did not see
anything obviously wrong. Are you starting spamassassin during boot?
> Even after removing gdm from the runlevel readahead still reads too much
> from the hard disk:
> (/etc/boot contains just some xfs stuff)
> Maybe readahead has some problems with a particular init script.
The only one I know about is preload, which will load all the files
used by the user after login at the end of the boot to speed up login
More information about the initscripts-ng-devel