Thanks for the patch! This one time, at band camp, Ralf Dragon said: > > /usr/share/doc/hdparm/README.Debian mentions why hdparm does not run > > it's init script if the array is rebuilding. I do not use software raid > > myself, so I am relying on reports from others. > > I grepped this file for "raid", sorry that I didn't read it completely.
No sweat - I have tried to clean it up and reorganize it a little, to make it easier to find stuff in it. > # default settings for hdparm > # this file is sourced from /etc/init.d/hdparm > > # uncomment this if you really want to start hdparm even if md is not ok > # <perhaps some explanations why this is not enabled by default> > #DIRTYRAID=OK I don't think this is something that you want to set on a permanent basis - hdparm should only be run on a dirty array if you have thought about it ahead of time, and decided to go ahead. > # <other variables> > DISC= > DEFAULT= > OPTIONS= > DEF_QUIET= > OPT_QUIET= These aren't really meant to be set directly - they are set to null initially just to make sure that they aren't accidentally imported from the environment. > ! if ! [[ "$raidstat" == 'OK' || "$DIRTYRAID" == 'OK' ]]; then > report_error "RAID status not OK. Exiting." > exit 0 > fi The '==' construct (as opposed to just '=' ) is not POSIX, and a number of shells choke on it. Just pointing it out - the script in general is not completely POSIX compliant, although I am slowly getting it closer. So: The only good argument to putting the bypass option directly into the init script via a defaults/ file that I can see would be if hdparm could then be run _before_ the array starts to rebuild. Unfortunately, array rebuilding starts even before /etc/rcS/07, so we buy nothing and gain the potential to really screw things up:(. I think the best solution here is to evalutate an environment variable (I have arbitrarily chosen FORCE_RUN just now) so the admin can bypass the simple-minded safety checks (and in this implementation, it skips even the evaluation of the kernel command line). This means running the init script by hand after boot up, but I think forcing this is OK, as it's not something that (IMHO, at least) should be encouraged to happen non-interactively. I have also documented this in README.Debian, as well as the safety checks themselves. Does this seem sufficient to allow you to do what you need to do? -- ----------------------------------------------------------------- | ,''`. Stephen Gran | | : :' : [EMAIL PROTECTED] | | `. `' Debian user, admin, and developer | | `- http://www.debian.org | -----------------------------------------------------------------
pgpKGCVmWzsiA.pgp
Description: PGP signature
