Date: Sun, 25 Apr 1999 18:03:50 -0500 From: David Kelly <dkelly@hiwaay.net> To: freebsd-scsi@FreeBSD.ORG Subject: Blocksizes (was Re: i/o error with larger QIC) Message-ID: <199904252303.SAA38889@nospam.hiwaay.net> In-Reply-To: Message from J Wunsch <j@uriah.heep.sax.de> of "Sun, 25 Apr 1999 15:06:47 %2B0200." <19990425150647.52666@uriah.heep.sax.de>
next in thread | previous in thread | raw e-mail | index | archive | help
J Wunsch writes: > = > . QIC only accepts three possible block size settings: 0 (variable), > 512, 1024. 8 mm accepts everything up to 240 KB. I don't know how > other tape systems behave here, but i haven't seen anybody who would > use odd fixed-length blocksizes either, so it's probably rather a > mood point. For sane normal Unix blocksize isn't a life and death matter. OTOH I've worked the past several years in a "data center" where tapes and other media was deposited and checked into the "library." Often we had no idea what was on the tape or how it was written. But if somebody requested a copy it was my job to figure out how to dupe it if the technician's cookbook approach failed. Most tapes were 8mm. Most tapes were in tar format. Most tapes were not blocked at 10k. A lot of tapes were apparently written on a VAX directly from FORTRAN. In this case the first block was 80 bytes. The next couple of blocks would be of random size, then a repeating pattern of blocksizes would emerge. Often (3) EOF's were used to mark a single intended EOF. Solaris provides a "Berkeley" tape device which eats these 3 EOF's. Otherwise the Solaris tcopy falsely stops early. I ported FreeBSD's tcopy to Irix and had one special version where I was hacking on the EOD detection. Got bit in the difference between Irix 6.2 and 6.3 here. Recently saw a tape with 488 tape files. Each with 3 EOF's. Took 8 or = 12 hours to tcopy as apparently an Exabyte 8505 takes a long time to = write an EOF. Given a choice I'll always choose Irix over Solaris. Especially when hosting a system for lusers. But the most annoying thing about Irix is that it defaults to 256k and 126k (or is it 128k?) blocksizes for 4mm and 8mm respectively. Not too awful as /var/sysgen/master.d/scsi lets you select the default. The bad thing is unless told otherwise the SCSI tape device driver will write a new tape with whatever blocksize was last seen. Can you say, "tar 2G of data with 512 byte blocksize?" Or = 31k? Or how about the luser who was writting 1024k blocksize because = the system would do it and he thought it would be faster? One contract ends, I get moved to another with similar duties but a chance to start with a clean sheet. So far we only have one system and its running FreeBSD with (4) Seagate Scorpion DDS-3 drives attached. Its serving well as I have more control of our incoming data now. OTOH the hardware wish list includes an SGI O2 simply the injest data the FreeBSD system can't grok. One day I'll experiment with the kernel parameter which limits max blocksize (have "HOWTO" archived) but probably won't get around to it until I'm under the gun. -- David Kelly N4HHE, dkelly@nospam.hiwaay.net =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D= =3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D=3D The human mind ordinarily operates at only ten percent of its capacity -- the rest is overhead for the operating system. To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-scsi" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?199904252303.SAA38889>