On 2019-04-19T10:32:14, Sitsofe Wheeler <sitsofe@xxxxxxxxx> wrote:
Yes, that's how this currently works.
Every IO is essentially translated to a separate object.
S3/Swift do not support overwrites, you can only PUT a single object as
a whole. (Multipart uploads are handled somewhat differently, but the
end result is the same.)
> > [create]
> > filename=/data.dat
> > rw=readwrite
> > bs=2m
> > size=50g
> > io_size=10m
> >
> > The PUT goes to /data.dat_0_2097152 signifying offset (0) and
> > xfer_buflen of 2M. Except, AWS seems to create the file
> > /data.dat_0_2097152. How do I make AWS understand this format?
> > Shouldn’t we be using RANGE requests?
"filename" is misleading. It's essentially just the prefix. The engine
is a bit of an exploit - it uses the offset/blocksize as an extension
and thus creates many separate objects.
What you want to do, S3/AWS can't do.
Regards,
Lars
--
SUSE Linux GmbH, GF: Felix Imendörffer, Mary Higgins, Sri Rasiah, HRB 21284 (AG Nürnberg)
"Architects should open possibilities and not determine everything." (Ueli Zbinden)
