Eric Sproul wrote: > Hi, > I've got an snv_99 host where the rpool mirror keeps showing checksum errors. > I've replaced both drives and it still occurs. It seems to be triggered by > sustained I/O on the pool, such as doing a zfs send or reading/writing large > amounts of data. > > The motherboard is a Tyan S2882 (Thunder K8S Pro), and the rpool disks (2 80GB > WD Caviars) are connected to the onboard SiI 3114 SATA ports. The bulk of the > system's storage is on 6 1.5TB SATA drives connected to an add-in Supermicro > 8-port card, and I have never seen errors on that pool. This leads me to > believe it is not a RAM issue (the RAM is also new, from a recent upgrade). > Has > anyone seen corruption caused by the SiI 3114 controller? The one thing I > have > yet to try is swapping the SATA cables, but this is a production server, so my > maintenance windows are limited. > > Below is a scanpci output. Any advice would be appreciated. >
If your not getting any other types of errors reported back (wierd driver errors, etc) I'd go for replacing everything you easily can (cables), retest, and then go from there. Following the cables, I'd opt to use a SATA card other than the 3114. Just a hunch, but I think thats your problem. On the more unusual side, check the power to the drives. It is possible, however unlikely, that the drives are getting insufficient voltage and misbehaving under high load... I doubt it, but you know, just have a check anyway. I assume when you say "rpool" you mean "remote pool", for backups perhaps? If thats the case, export the pool, swap the cables, import the pool and give it a shot, that shouldn't require a production outage, just do it in a time when any badness doesn't impact production load. Also... have you asked FMA what it thinks? "fmadm faulty -v" In my experience FMA rarely actually catches hardware issues but none-the-less sometimes it gets it right, so give it a check. benr. -- Ben Rockwood cuddletech.com Joyent Inc. PGP: 0xC823A182 @ pgp.mit.edu "...even at night his mind does not rest. This too is meaningless." - Ecclesiastes 2:23 _______________________________________________ storage-discuss mailing list storage-discuss@opensolaris.org http://mail.opensolaris.org/mailman/listinfo/storage-discuss