SCSI Log Sense -- Error Counter Pages

[Richard Bickers · Accepted solution]

OK, I got it.

I was getting mixed up between datasets and CCQs. The capacity loss calculation is:

Re-written_CCQs/Good_CCQs * 100%

The key point is that in LTO 3, there are 128 CCQs per dataset (different figures for LTO 1 and 2).

So in your example, you'd written 24.5GB which is 15,300 datasets - which is 2M CCQs. In that time you'd re-written 14,000 CCQs which works out as 0.7%. Anything less than 1% is very good.

If you want to track capacity loss on the fly you can use:

Param_4/(Param_5 * 128) * 100%

General rule of thumb:
- <1% very good
- 1-5% normal
- 6-10% less normal but don't worry
- 10-20% maybe clean the heads - will probably self-clean anyway
- >20% we start to worry if it stays up here. Think contamination.

Sorry for the confusion. I got there in the end...

Good luck with the rest of your testing.

[Brian Eickhoff]

One more piece of needed data...

What is the number of CCQs per dataset for LTO 2?

Thanks!

[Richard Bickers]

No worries. There are 64 CCQs in a dataset for LTO 2. Also true for LTO 1. We made the change for LTO 3 as the datasets are bigger.

Cheers,

Richard.

[Brian Eickhoff]

Have you learned anything more about what the numbers represent in Parameter 0? The ones that are "not used"? Just wondering...

[Richard Bickers]

I have. Parameter 0 contains the number of datasets that had a CCQ re-write in - i.e. one or more. That's why Parameter 0 tends to be a subset of Parameter 4.

We're interested in the total number of CCQ re-writes rather than this subset so the value isn't all that useful.

I've got a full run down from the firmware team on all of these parameters now. The rest of them are even more obscure... Shows what happens when a generic SCSI spec is left to interpretation!