[Soekris] openbsd compact flash boot and diskless run scripts

Chuck Yerkes chuck+soekris at 2003.snew.com
Sat Dec 13 05:56:14 UTC 2003


Quoting Gray Watson (gray.soekris at mailnull.com):
> > What's the advantage to putting all of you /usr/ into RAM, when
> > they usually ARE in RAM - running - if your using them.
> 
> For me its the general ability to treat the system as a semi normal
> one with shell logins, scripts, logs, etc. that would kill the CF fast
> -- that is if I read the web opinion on their read/write cycles.

Then, er, don't.

My bloated setup is on a 128 (the small one is on a 16).

I log in all the time.  /dev/ must be writable.  I have it on a (170k)
MFS.   /mfs (mentioed before) is on a 500k partition.

I have syslog of everything off to another machine and to a
/tmp (8MB MFS) file that newsyslog rotates at 1MB.  I use that
only when I'm trying something new and am being lazy. 

Scripts?  I don't really write on the system.

Reads you can generate all you want.  try it.  Rumors of death of
a sector are at 1 million writes...  1 million writes is a lot of RCS
checkins.  (and those rumors may be questionable with currect CFs).
Swap will kill you, but rsync'ing a new /usr up every couple months
won't.

> I don't want to treat the box like something I can't ever ssh into
> without worrying how many reads I'm generating or how this or that
> application deals with the filesystem.

Well, you SHOULD know.

ssh will cause a lot to be sysloged (off the box in a reasonable
situation, or /dev/null is you live on the edge).  There's a chmod/chown
of the pty and... well that's it.  Your shell is read and . files.
Read.  Only.

My 4501 feels no slower than my sparc 20.

Vs running out of RAM which will stop things rather suddenly.

> I got about ~10mb free in the MFS and another ~10mb in core.  That's
> all I need but YMMV.



More information about the Soekris-tech mailing list