DragonFly BSD
DragonFly users List (threaded) for 2005-12
[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]

Re: Waiting 15 seconds for SCSI devices to settle *groan*

From: Bill Hacker <wbh@xxxxxxxxxxxxx>
Date: Sun, 04 Dec 2005 00:39:27 +0800

Hiten Pandya wrote:


What we really need is a loader option that would control the SCSI probe delay. If I recall correctly it is still a compile time option. Another thing that we can offer is a boot menu option that would cut shortcuts exactly like these across various subsystems where applicable for people who do not have "delicate" hardware.


There has long been a workaround, and it is a very good one, *IF* the devices in question are not the boot devices, which large storage arrays seldom are:

- get the box on its feet and ssh'able.  Leave the non-boot mass-store
behind so as to do that quickly.

- run a script that does a CAMCONTROL or ATACONTROl scan/reinit of the
devices not handled at boot time (if need be).

- fsck these *after* the system is at least 'communicating.

Depending on what the 'left for later' devices are used for, delay
starting services that need them.

I do this at both startup and shutdown with all the boxen on which I
have large secondary and tertiary ATACONTROL IDE RAID1, as I want a full
fsck, not a preen, at both ends of their time on-watch, and want quicker
connectiviy to the box that fsck'ing multiple 200GB+ RAID arrays would
otherwise allow. (15 minutes for some boxes)

No harm if PostgreSQL, Zope/Plone, rsync, or Exim have to wait, so long
as I have ssh right away.

This also reduces the risk of a remote machine 'hanging' and out of ssh
contact, or of contention over a background' fsck and some app that
wants what it cannot yet have.


Bill Hacker

[Date Prev][Date Next]  [Thread Prev][Thread Next]  [Date Index][Thread Index]