minfree 1 -> 0 -> 1 == death ... PLEASE HELP

[Eric Anderson]

Juri Mianovich wrote:
> On Mon, 27 Aug 2007, Eric Anderson wrote:
> 
>> Regarding the 'KABOOM' part - is that when you mount
> it?  How about RO?
> 
> 
> No - it mounts and runs just fine.  For about an hour.
>  Once any meaningful activity is run on it for an hour
> or so, the aac raid controller dies, spewing
> "controller is no longer running" messages on console.
> 
> If I skip that filesystem, and mount the other
> partitions, the system will stay up indefinitely.  It
> is only by running that filesystem (that I changed
> from 1 to 0 to 1) that the aac controller will die
> off.
> 
> 
>> Have you fsck'ed the fs at all?
> 
> 
> Yes.
> 
> 
> 
>> Also - please send the output of this command:
>>
>> dumpfs /dev/aacd0s1e | head -n 40
> 
> 
> Here you are - sorry for the bad linewrapping:
> 
> 
> magic   19540119 (UFS2) time    Mon Aug 27 19:28:15
> 2007
> superblock location     65536   id      [ 44967b53
> b7c98a12 ]
> ncg     10379   size    976478879       blocks 
> 945756917
> bsize   16384   shift   14      mask    0xffffc000
> fsize   2048    shift   11      mask    0xfffff800
> frag    8       shift   3       fsbtodb 2
> minfree 1%      optim   space   symlinklen 120
> maxbsize 16384  maxbpg  2048    maxcontig 8    
> contigsumsize 8
> nbfree  1520861 ndir    1336402 nifree  233593652     
>  nffree  1151519
> bpg     11761   fpg     94088   ipg     23552
> nindir  2048    inopb   64      maxfilesize    
> 140806241583103
> sbsize  2048    cgsize  16384   csaddr  3000    cssize
>  167936
> sblkno  40      cblkno  48      iblkno  56      dblkno
>  3000
> cgrotor 8911    fmod    0       ronly   0       clean 
>  1
> avgfpdir 64     avgfilesize 16384
> flags   soft-updates
> fsmnt   /mount2
> volname         swuid   0
> 

[..snip..]

> Like I said, I am sure there is a fascinating
> explanation for all of this and we can all learn a
> lot, but I _don't care_.  Why is the aac controller
> dying ?  Don't care.  Why can't the system handle a
> minfree of 0 ?  don't care.  Why does my new minfree
> of 1 behave like a minfree of zero ?  Don't care.
> 
> All I want to know is, how do I get back the old
> minfree of 1 I had 24 hours ago instead of the "new
> and improved" minfree of 1 that I have now ?

Can you try doing:


tunefs -o time /dev/aacd0s1e

and then mounting it?


Eric