[TuxOnIce-devel] Please hammer 3.0.99.40

[Nigel Cunningham]

Hi.

Kenneth Crudup wrote:
> On Fri, 11 Dec 2009, Nigel Cunningham wrote:
> 
>> With last night's release of 3.0.99.40, I believe I fixed the problem
>> that Kenneth had with occasional resume failures.
> 
> Heck, or even before that:
> 
> ----
> TuxOnIce debugging info:
> - TuxOnIce core  : 3.0.99.32
> - Kernel Version : 2.6.32-TOI-GIT-NC
> - Compiler vers. : 4.4
> - Attempt number : 10
> - Parameters     : 0 667656 0 1 -2 0
> - Overall expected compression percentage: 0.
> - Compressor is 'lzo'.
>   Compressed 3859136512 bytes into 1604905995 (58 percent compression).
> - Block I/O active.
> - Max outstanding reads 4767. Max writes 2337.
>   Memory_needed: 1024 x (4096 + 336 + 112) = 4653056 bytes.
>   Free mem throttle point reached 4219.
> - Swap Allocator enabled.
>   Swap available for image: 2268510 pages.
> - I/O speed: Write 136 MB/s, Read 181 MB/s.
> - Extra pages    : 1324 used/2048.
> - Result         : Succeeded.
> ----
> 
> And I know it's hitting both the SSD and the HD, 'cause that I/O Speed
> figure is about a third of what it should be. I almost kinda hate to
> bring the box down to run the new drops! :)

Yes, but it wasn't whether it was using both that was causing you to see
the issue, Kenny. It was swap usage, particularly swap usage such that
the first page available for TuxOnIce at hibernate time was isolated
from the next page, so that the were in separate extents. In short, you
will still hit that bug sooner or later with 3.0.99.32 - it's just a
matter of time.

Regards,

Nigel