Date: Wed, 5 Mar 2008 17:55:00 -0500 (EST) From: Daniel Feenberg <feenberg@nber.org> To: Kevin Kinsey <kdk@daleco.biz> Cc: Bill Moran <wmoran@potentialtech.com>, freebsd-questions@freebsd.org Subject: Re: faster booting Message-ID: <Pine.GSO.4.64.0803051751580.5952@nber5.nber.org> In-Reply-To: <47CF2246.5050007@daleco.biz> References: <Pine.GSO.4.64.0803051450540.18940@nber5.nber.org> <20080305154351.fc53a07b.wmoran@potentialtech.com> <Pine.GSO.4.64.0803051623430.29394@nber5.nber.org> <20080305171739.1f51a11a.wmoran@potentialtech.com> <47CF2246.5050007@daleco.biz>
next in thread | previous in thread | raw e-mail | index | archive | help
On Wed, 5 Mar 2008, Kevin Kinsey wrote: > Bill Moran wrote: > >>>>> So, is there advice anywhere about speeding up the boot process? It >>>>> appears that most of the 1 minute 45 seconds to boot our system is wait >>>>> time for checking the existence of non-existant hardware and would not >>>>> be >>>>> appreciable reduced with a faster CPU or disk. Are there kernel options >>>>> that we could use to avoid this checking? Would recompiling the kernel >>>>> in >>>>> some specialized way help? Would pico-bsd be faster? >>>>> >>>>> About the only thing I can find is to reduce the 10 second boot screen >>>>> delay - but we need to cut more than 30 seconds. >>>>> >>>>> The server is statically configured but the clients obtain network >>>>> configuration from dhcp and pxeboot with nfs mounted root directories. >>>>> Clients are FreeBSD and Linux, and we are not eager to give up pxeboot >>>>> as >>>>> it has greatly simplified maintainance. >>>>> >>>>> Any suggestions, pointers much appreciated. >>>> Three things I can think of: >>>> * The 10 sec boot delay, which you already mentioned >>>> * Make sure the wait time for SCSI devices is a low as reliably works. >>>> If it only has SCSI disks, this could probably very short, 1 sec or so >>>> * Recompile your kernel removing any devices that don't exist in your >>>> hardware. >>>> >>>> I'm not buying this, however. My laptop boots in ~30 seconds with a >>>> mostly stock kernel. Please provide specific details as to what's >>>> slowing it down. Are you sure it's not a slow BIOS? Many of the Dell >>>> systems we have take several minutes with BIOS self-checks before the >>>> OS even starts to boot. >>> The BIOS time isn't terrible - BTX shows up on the console within 15 >>> seconds. The major delays happen when the last console message is about >>> atapci: (25 seconds) and ad2: (15 seconds). >> >> Funky. That's a Looong time to wait for an ATA controller to determine >> whether or not their's a disk attached. Do you have an ad2? If not, >> you might want to check the BIOS to see if there's an option to disable >> that particular part of the ATA chain to see if that speeds FreeBSD's >> probe up. > > Let's be sure of this, though; are we actually talking about an ATA > controller issue? The phrase "last console message" doesn't necessarily > mean it's the ATA controller, but whatever is *next* in the bootup process, > AFAICT, *after* the probe of /dev/ad2, which, on my systems > is the mounting of the root filesystem. Yes, there is an ad2 - it is the root filesystem, but given the point made above, it might be that the best thing to do is put that on a faster device. It is currently on a 2.5" drive that was selected to reduce power consumption and make the UPS last longer. Maybe a thumb drive would be better. As for the suggestion that we delay the clients, we plan to enable memory testing in the BIOS of the clients to delay the first request for dhcp services. Any delays placed later in the boot sequence won't help with the problem. Dan Feenberg > > OTOH, turning off BIOS probes for disks that don't exist is > a good idea, IMHO. > > Kevin Kinsey >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.GSO.4.64.0803051751580.5952>