ATA problems again ... general problem of ICH7 or ATA?

Miroslav Lachman 000.fbsd at quip.cz
Sun Aug 20 19:13:30 UTC 2006


Greg Byshenk wrote:

> On Sun, Aug 20, 2006 at 07:51:29PM +0200, Miroslav Lachman wrote:
> 
>>Greg Byshenk wrote:
> 
> [...]
> 
>>>This happened four times (with the same errors that have been discussed
>>>here), running 6.1 STABLE as of June 22.  Before attempting to RMA the
>>>drives, I tried an updated kernel, 6.1 STABLE as of July 19.  Strangely
>>>enough, the problems disappeared.
> 
> 
>>>So, while I have not checked everything that has changed, it _might_ be
>>>worth trying 6.1 STABLE...
> 
>  
>>I have problems with 6.1-RELEASE same as with 6.1-STABLE from August 2. 
>>I can try newer STABLE, but as I see on cvsweb, there are not much 
>>changes in ATA driver sources, only new chipsets added.
> 
> 
> It is only an idea, based on something that worked for me.  And, as I
> said, my situation is not exactly the same as the others.
> 
>>It is strange to me, that I can see significant changes of read/write 
>>speed. (I am running nonstop tests with writing disk full of files, 
>>delete them, and start again + generating graphs) Speed vary from 
>>2.5MB/s to 11MB/s by jumps. Not continuous from the lowest to the 
>>highest. Writing is for example 3MB/s for 20 hours, then jump to 10MB/s 
>>and after some time (6 - 20 hours) jump down to about 3MB/s.
>>After some days of testing, disk disappear, system reboots itself, 
>>resynchronize gmirror and work for next few days till the next disk lose.
>>Also earlier synchronization was done after 1:30 hour (at about 30MB/s), 
>>now synchronization run at lower speeds - from 2.5MB/s to 15MB/s, so the 
>>whole synchronization is done after more then 5 hours (the longest was 
>>20 hours to synchronize 250GB HDDs)
> 
> 
>>I don't know what more can I test, what more could be done to solve 
>>these problems. :(
> 
> 
> You are using gmirror, which I am not, so the situations are not
> analogous, since my situation was with h/w RAID.  And I have no direct
> experience with gmirror (I use gvinum on a couple of secondary systems,
> but those are SCSI based).
> 
> Does the output of 'systat -vm' tell you anything of interest?  That is,
> are the disks running at or close to 100%, are the CPUs fully loaded, or
> anything else...?

There is nothing interesting in systat / gstat / top or anything else.
System is almost idle, just running test script for disk writing. Speed 
problems is not dependent on gmirror. I deactivated gmirror on second 
machine and run test on normaly mounted filesystems with same low speeds ;(

This is systat from gmirrored system running test:

     4 users    Load  0.01  0.02  0.00                  Aug 20 21:06

Mem:KB    REAL            VIRTUAL                     VN PAGER  SWAP PAGER
         Tot   Share      Tot    Share    Free         in  out     in  out
Act  124144    9580   897616    27168   43016 count
All 1016728   75888364464876   210508         pages
                                                                  Interrupts
Proc:r  p  d  s  w    Csw  Trp  Sys  Int  Sof  Flt        cow    4144 total
            7 75       979    4  155  288   27      161484 wire 
1: atkb
                                                    191292 act 
14: ata
  0.4%Sys   0.0%Intr  0.0%User  0.0%Nice 99.6%Idl   624272 inact    11 
16: bge
|    |    |    |    |    |    |    |    |    |      41360 cache   133 
19: ata
                                                      1656 free   2000 
cpu0: time
                                                           daefr  2000 
cpu1: time
Namei         Name-cache    Dir-cache                     prcfr
     Calls     hits    %     hits    %                     react
         6        6  100                                   pdwake
                                           zfod       1354 pdpgs
Disks   ad4   ad6                         ozfod           intrn
KB/t    125   126                         %slo-z   113888 buf
tps      34    33                    1407 tfree        17 dirtybuf
MB/s   4.13  4.10                                   69977 desiredvnodes
% busy   54    48                                   20661 numvnodes
                                                     17286 freevnodes



More information about the freebsd-stable mailing list