I have two AIT-2 tapes which have been mounted 467 times, and one LTO
tapes which is at 272. The remainder of the LTO's are below 100 mounts,
with most below 30 mounts. Amazingly, the AIT-2 tapes have survived a
drywall construction project in the datacenter, and four questionable
tape drives. These used to be the catalog when we were only AIT, and
now they don't do anything. Thankfully, I've never had to read them for
a media server recovery, and when I inspect the logs, I do not see media
errors on these tapes. In my battle with bad media (caused by drywall
dust) I have seen tapes write with zero errors, and then fail completely
on the read. I've also seen these same tapes fail on the first attempt,
but succeed with some errors on the second attempt. I conviced
management that we cannot trust ANY data written on the AIT-2's, so now
I have a small qualstar library and about 700 tapes to play with. Our
ADIC 10K has been an excellent performer.
I'm curious to know if anyone has seen a pattern of LTO failures
attributed to a high mount number.
-Jon
> It's all guesswork. Seriously, I don't think you'll ever get a solid
> recommendation until NB keeps more usage data. Vendors can tell you
> number of "full pass writes" a tape is rated for and that sort of
> thing, but how do I translate that into NetBackup's max mounts? In
> NB, a mount for write is a mount for write, whether it writes 1MB or
> 100GB.
>
> NB usually does a pretty good job of filling up tapes and if you have
> lots of clients running at the same time with multiplexing, etc. it's
> likely your drives will mount a tape and write it through to the end.
> But what if the data flow tapers off? Shoe-shining is going to stress
> the cartridge a lot more than streaming. And perhaps you later mount
> that tape for a few small backups...you see how difficult it is to
> keep track of this ;) I think NB *could* keep this data...(psst,
> Veritas guys, this would be a cool feature)
>
> We use 125 mounts for LTO (which is also what we used for DLT). Maybe
> I could get more mounts. I suspect I could. But the value of the
> data is nearly infinite in comparison to the cost of the tape, so this
> decision is easy. On the one hand, maybe I spend too much for tapes.
> On the other hand, I'd rather do that than lose some critical
> enterprise data ("But boss, I saved $50!") Even with multiple in-line
> copies, off-siting for DR, etc. I still feel that way ;) Gaaaa - I
> hate this issue because I can't stand throwing away tapes that might
> still be serviceable but how do you know...
>
> Some of the STK drives (99xx) use 300 max mounts. IBM's 3590s use a
> higher number as well.
>
> Out of curiosity, what do you use for number of hours to clean? I use
> 300. I haven't noticed any correlation between the (few) media
> problems we've had and cleaning hours.
>
> I like your idea of using "old tapes" for development or non-critical
> data. I hadn't thought of that.
>
>
> -----Original Message-----
> *From:* Mark Eisenhardt [mailto:Mark_Eisenhardt AT stoneybrookfl DOT com]
> *Sent:* Wednesday, December 04, 2002 5:26 PM
> *To:* White, Steve; 'Jonas Blåberg'; veritas-bu AT mailman.eng.auburn DOT
> edu
> *Subject:* RE: [Veritas-bu] Max number of mounts for an LTO tape?
>
> Just thought I'd throw my 2 cents in on this because I've never
> heard of any rule of thumb on the number of mounts/usage
> for tapes.
> I've seen some other backup software that sets the default to 360.
> ie Omniback
> For Netbackup we set our production pool to 120 mounts (Business
> Critical Data)
> The development pool is set to 360 mounts (Non-Critical Businesss
> Data)
> Also we have been on Veritas NBU for 1 1/2 years and have yet to
> exceed 50 uses per tape, and no tape failures, so it will be a
> while before we can establish an SOP for this.
>
> Mark
>
>
|
