doug> sigh, it has been years since I tried-and-failed.... Time to try again? Linux has certainly advanced in the interving years.
Doug> putting a linux-based NFS server into a high-traffic engineering Doug> farm. (pesky HW engineers and their simulation or design data). Ask your boss how expensive the time those engineers spend doing simulations, vs. the cost of downtime. Then you'll be able to justify more money on reliable storage. Doug> I ended up putting freeBSD on the $10k IDE-raid box I bought in Doug> 2001. Can't do the freeBSD thing for a "new requirement" on a Doug> LOWER!? budget, so: This make no sense. FreeBSD is free. Where's the cost? Doug> THE QUESTION: Now+again, I wonder who has tried and Doug> failed/succeeded with cheap (linux) NFS server... ? You're looking to save money in all the wrong place IMNSHO. I too work with hardware engineers doing ASIC chip simulations. Downtime or data loss is WAY WAY WAY more expensive than $20k or even $100k for us. If an engineer's 3 day simulation craps out with 20 minutes left to run (shades of Hitchkikers Guide to the Galaxy) because the NFS server died, then he's lost those three days, and turn tape out, it's a big issue. Heck, just look at the license costs for a single user license from Cadence, your $20k is noise I tell you noise! :] Doug> ok, to do what? Get a netapp. Doug> oh, just read a few and write a lot of GB in parallel Doug> (multi-NFS-clients). e.g. lots of files (100k here, 100k there Doug> per hour...stay the course...) Doug> Is any linux distro better than another for this? RHEL ? (or Doug> whiteboxEL which works the same, right?) other? I'd try CentOS 4 (maybe 5 now) and then talk with Trond who's the NFS maintainer. Also, you don't mention what clients you have running either. Doug> Don't need to reminded of obvious BETTER solutions ... I'm Doug> trolling for "what works great new-and-under-$5k". Your boss is being stupid. Ask him how much is costs for an engineer to sit around on their ass for a day while the NFS server gets fixed. Then factor that cost in against the cost of a NetApp box. Doug> The recommendation I'm trying to fight for= NetApp. Cmon, $20k Doug> is nothing!) Yup, you're totally right. Doug> Sorry this is long. thanks for listening/replies. Gory detail below. Doug> How did it fail me on (redhat-before-FC; tried 2.5 kernel, no Doug> better...)? Despite giga-memory and excess I/O avail for Doug> enet+disk, what puked was NFS service. I bumped up nfsd Doug> #processes, tried various wsize,rsize, and sure it'd WORK, but Doug> add a few more busy NFS clients, and the server seemed to Doug> self-destruct. Symptoms: 20+ load, crawling NFS response. NFS over TCP, UDP? NFSv2 or NFSv3? Client OS? Doug> Sorry, my "requirements" and "symptoms" are imprecise, but it Doug> only seemed to take a few hours and a few McWhoppers, i mean a Doug> couple gb mostly-written by a couple users IN a few hours, to Doug> CHOKE/gag/pretty-much-hang Linux. Havent bought RHEL...too wary Doug> I am? RH/RedHat (nfs client!) systems do "play nicely" Doug> otherwise here, though. Network thruput can be 20gb-30gb/hour Doug> on 100mb enet, IF you don't use NFS. Doug> What DOES work (NFS server) without choking, in my budget+history, has been: Doug> --NetApp (which I still recommend. Boss says $18k==toomuch). Doug> (I say: a steal! Price has halved every 3? yrs!) Doug> --freeBSD (not the perfect fit either for a Doug> remote-site-main-"filer") Because you don't have serial console access to the box? Doug> Here's a possible HW platform, if the OS loaded onto it was appropriate: Doug> Dell PE1900 poweredge w/optional integrated SAS/SATA raid5, Doug> 5x500gb SATA 7.2krpm, probably close to 2tb usable, Depends on the filesystem. Me, I'd go with ext3 on there, maybe JFS. Stay away from Reiserfs3, and reiserfs4 isn't out of beta yet either... XFS has some potential, but I dunno... haven't seen great words about it. Anyway, with the filesystem, and RAID5, you'll get more like 1.7Tb of useable space. Not bad. Doug> Broadcom NetXtreme gb-nic (something I havent GOTTEN/had Doug> yet), (2) 2.0ghz dual core xeon 5130 4mb-cache, 1333mhz Doug> FSB, Doug> $2900+tax+ship=$3200 or $3300. No OS. (OS is Doug> my question/problem, or you could say lack of budget for Doug> an $18k 1+TB NetApp is my problem.) System looks fine to me. Though you don't mention memory. In any case, for an NFS server you shouldn't need that much memory. Doug> Perhaps this troll will perk up my energy/luck. If given a good Doug> tip, I WILL try free ideas/OS/tweaks first :) Impossibly, I must Doug> figure out if it will work WELL before buying anything. >From reading your info, you need to give us more details. And if you have a spare box lying around, why don't you try and setup CentOS4 (or 5) on there and point your clients at it for a test. Then you can test out things without breaking your production NFS server if things go south. John _______________________________________________ bblisa mailing list [email protected] http://www.bblisa.org/mailman/listinfo/bblisa
