vBulletin Search Engine Optimization
| |||||||
| Register | FAQ | Members List | Calendar | Search | Today's Posts | Mark Forums Read |
| ||||
| Hi all, I've got a bunch of questions regarding performance and upgrade options for disk subsystems on our database boxes, and I wondered if anyone would be able to shed any light on compatibility, certainly with some of the new storage arrays, and possibly our old boxes too. So, thanks for your time! Background: Database machine running Sybase (32-bit; bleh) w. 4 engines on 4 procs on an E4500 (Solaris 8) Storage is a couple of A5200s (fully populated with 18GB and 73GB) mirrored in Veritas. Unfortunately the DB is big at ~250GB. Our current version of Veritas is no longer supported and we've been having lots of small niggling problems with the arrays over the years (loops disappearing, arrays very occasionally locking up) so we're begining to thing about upgrade/new storage options. Leaving aside the option of upgrading our Veritas and trying to solve the array issues if we were to go for new storage, it looks to me like there are two options, namely new SCSI or new fibre On the FC side, there's the 3510, are these definitely backwards compatible with our FC host cards? what are people's experiences with them? I've read mixed reports on DejaGoogle already. The other solution would be a 3310 as a) SCSI arrays have got faster, cheaper and bigger and b) we've had these problems with the fibre loops. The question here is whether or not we can add a suitable SCSI adaptor, and whether the PCI bus would be fast enough. Part of me would like to part exchange them all and look at something like a V490 with a couple of the new arrays; but the likelyhood of spending $200k+ on a live and failover box is slim If anyone who's had any experience with these newer arrays (especially running them in an older environment) has any pointers, I'd be very greatful. Thanks! |
| |||
| * Steph wrote: > Our current version of Veritas is no longer supported and we've been > having lots of small niggling problems with the arrays over the years > (loops disappearing, arrays very occasionally locking up) so we're > begining to thing about upgrade/new storage options. Leaving aside the > option of upgrading our Veritas and trying to solve the array issues if > we were to go for new storage, it looks to me like there are two > options, namely new SCSI or new fibre This isn't answering your question, but we've had a fair amount of excitement with A5200s, and there are two issues which have made things better: make sure the disk firmware is right, and beware of taking disks out of some slots in the arrays. Wrong disk firmware seems to be able to cause intermittent total loss of contact with the array, and removing disks from the wrong slots can cause weirdo loop flapping problems. --tim |
| |||
| On Thu, 27 Jan 2005 17:14:53 +0000, Tim Bradshaw <tfb@cley.com> wrote: >* Steph wrote: > >> Our current version of Veritas is no longer supported and we've been >> having lots of small niggling problems with the arrays over the years >> (loops disappearing, arrays very occasionally locking up) so we're >> begining to thing about upgrade/new storage options. Leaving aside the >> option of upgrading our Veritas and trying to solve the array issues if >> we were to go for new storage, it looks to me like there are two >> options, namely new SCSI or new fibre > >This isn't answering your question, but we've had a fair amount of >excitement with A5200s, and there are two issues which have made >things better: make sure the disk firmware is right, and beware of >taking disks out of some slots in the arrays. Wrong disk firmware >seems to be able to cause intermittent total loss of contact with the >array, and removing disks from the wrong slots can cause weirdo loop >flapping problems. > >--tim Any docs from sun that point to the slots that may give trouble? I've had wierd loop problems on my A5x00's before as well. It'd be great to know which slots were likely trouble makers. |
| ||||
| On Fri, 28 Jan 2005 20:08:33 +0000, Allen Wooden wrote: > On Thu, 27 Jan 2005 17:14:53 +0000, Tim Bradshaw <tfb@cley.com> wrote: >>This isn't answering your question, but we've had a fair amount of >>excitement with A5200s, and there are two issues which have made >>things better: make sure the disk firmware is right, and beware of >>taking disks out of some slots in the arrays. Wrong disk firmware >>seems to be able to cause intermittent total loss of contact with the >>array, and removing disks from the wrong slots can cause weirdo loop >>flapping problems. >> >>--tim > > Any docs from sun that point to the slots that may give trouble? It's supposed to be 0,5 & 10 front, and 0,3,6 & 10 rear; something to do with signal degradation through the array; It's frustrating; we're losing more than a disk a month to failures on our 5200s (Four full arrays FWIW; so we're still losing over the odds especially as not all of our disks on those arrays are terribly active); hell, we lost an 18GB drive today when the 73GB next to it was replaced due to failure; and these two weren't even in the same disk group, or, for that matter, imported on the same server. Contrast that with an A1000 on an E3500 that's been rock solid for us. I've looked at our array firmware and everything is uptodate; though OS patches are somewhat lacking; I've not yet looked at disk firmware. I don't know; the flexibility of the Veritas is nice; but part of me's looking at A1000s doing mirroring in combination with SDS to replace the fibre channel; if we can't upgrade arrays due to compatibility (or manglement approval for budget) what's wrong with extra A1000s from eBay and brand new disks... |