Messages in this thread | | | Date | Fri, 12 Apr 2002 12:28:45 +0200 | From | Jochen Suckfuell <> | Subject | Re: sard/iostat disk I/O statistics/accounting for 2.5.8-pre3 |
| |
On Fri, Apr 12, 2002 at 01:06:10AM +0200, Zlatko Calusic wrote: > > major minor #blocks name rio rmerge rsect ruse wio wmerge wsect wuse running use aveq > > 3 0 10037774 ide/host0/bus0/target0/lun0/disc 2381496 3602026 12295120 24796300 2167401 11887863 28592436 168474630 -1 849341880 -666064880 > > > Also, it _seems_ that only on whole devices ios_in_flight variable > (column "running" in /proc/partitions) drops to -1. Your (and mine) > output shows it clearly, yes. This needs further investigation, I have > been unable to find the problem so far.
I have seen ios_in_flight as low as -4 already.
> And, as aveq value is calculated directly from the ios_in_flight > value, the average queue value is wrong too (it never goes negative, > if everything is all right). The use (one of the most important values here) is also completely wrong when ios_in_flight is too low: Whenever ios_in_flight!=0, the disk is accounted as running.
I have patched gen_partition_disk() to correct the ios_in_flight value as a workaround. This is better than checking it each time ios_in_flight is decremented, because it only happens when someone reads /proc/partitions, and so it doesn't slow down the disk access.
But I'm still trying to reproduce the cause of the problem.
Bye Jochen
-- Jochen Suckfuell - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.kernel.org More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
| |