Date: Tue, 31 Jul 2001 12:19:21 -0400 From: Jay Sachs <jay@eziba.com> To: m p <sumirati@yahoo.de> Cc: freebsd-questions@freebsd.org Subject: Re: NFS v3 server locking fails Message-ID: <3B66DA89.C5E35C2@eziba.com> References: <20010731160100.78020.qmail@web13301.mail.yahoo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
m p wrote: > > > We're trying to get an Oracle db running on Solaris 2.6 to work with a > > 4.3-Stable (7/24) system as the underlying filesystem via NFS. The > > -Stable box has an ATA raid subsystem. The problem is that over NFS v3, > > the Solaris system fails to acquire the lock, and NFSv2, though it > > works, isn't really an option since it's (a) an order of magnitude > > slower and more importantly, (b) we have files > 2gig. > Why want you to do this? My experience says: DONT. And if you think about costs > (using an IDE-raid): DONT. > The problem is with the Oracle db that it is not nearly fast enough with NFS > (have done only little tests with NFSv3; sure about NFSv2). Another point is: > How do make sure that your network connection is permanent? > Have you tested recovery of a database where network-outage happend in the > middle of the update (eg pull the plug)? It is hard enough using only the local > filesystem. This isn't my choice, I'm just trying to deal with what others have (mistakenly, IMO) chosen. Your points are all reasonable. One thing is that a very similar box running Linux worked with acceptable speed under NFSv3. (The disks died due to insufficient cooling; that's been addressed.) The cost of the IDE raid apparently was very cheap (less than half) the cost of local SCSI disks on the Sun. One other datapoint is that the idea is to share this storage among several Sun boxes running oracle, and hence the decision to go with an shared NFS solution, instead of buying SCSI disks for each sun. These are not production data, but test and staging data, so the risks are mitigated somewhat. > Side-question: Why do you use Oracle databasefiles > 2gb? If you not have data > per tablespace in the three-digit GB range, go with smaller ones. On the most > (in this case HP-UX) you will get some percent more performance .. and a little > admin overhead. There is a major db re-partitioning scheduled in the future (3 months out), and the plan already is to reduce the filesize. Even if we were going to do this immediately, we have a chicken & egg problem: the files are already > 2gig, and we need NFSv3 to access them in order to do the repartitioning. > > My solution would be: Take some money and buy for the Sun more/bigger > SCSI-discs. Unfortunately, the money is already spent. The alternative is to switch back to Linux and use what I'm told are beta drivers for the RAID. jay To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-questions" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?3B66DA89.C5E35C2>