mirrors/git-annex
2
0
Fork 0
Code Issues Pull requests Projects Releases Packages 1 Wiki Activity Actions

finish multipart support using unreleased update to aws lib to yield etags

Browse source 
Untested and not even compiled yet.

Testing should include checks that file content streams through without
buffering in memory.

Note that CL.consume causes all the etags to be buffered in memory.
This is probably nearly unavoidable, since a request has to be constructed
that contains the list of etags in its body. (While it might be possible to
stream generation of the body, that would entail making a http request that
dribbles out parts of the body as the multipart uploads complete, which is
not likely to work well..

To limit this being a problem, it's best for partsize to be set to some
suitably large value, like 1gb. Then a full terabyte file will need only
1024 etags to be stored, which will probably use around 1 mb of memory.
This commit is contained in:
Joey Hess 2014-11-03 15:53:22 -04:00
parent 6e89d070bc
commit 8faeb25076
2 changed files with 19 additions and 8 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

19
Remote/S3.hs
View file

@ -13,6 +13,10 @@ module Remote.S3 (remote, iaHost, configIA, iaItemUrl) where
import qualified Aws as AWS
import qualified Aws.Core as AWS
import qualified Aws.S3 as S3
#if MIN_VERSION_aws(0,10,4)
import qualified Aws.S3.Commands.Multipart as Multipart
import qualified Data.Conduit.List as CL
#endif
import qualified Data.Text as T
import qualified Data.Text.Encoding as T
import qualified Data.ByteString.Lazy as L
@ -170,7 +174,7 @@ store r h = fileStorer $ \k f p -> do
multipartupload sz k f p = do
#if MIN_VERSION_aws(0,10,4)
let info = hinfo h
let objects = bucketObject info h
let object = bucketObject info h
uploadid <- S3.imurUploadId <$> sendS3Handle' h $
(S3.postInitiateMultipartUpload (bucket info) object)
@ -180,14 +184,13 @@ store r h = fileStorer $ \k f p -> do
, S3.imuExpires = Nothing -- TODO set some reasonable expiry
}
-- TODO open file, read each part of size sz (streaming
-- it); send part to S3, and get a list of etags of all
-- the parts
etags <- sourceFile f
$= Multipart.chunkedConduit sz
$= Multipart.putConduit (hawscfg h) (hs3cfg h) (hmanager h) (bucket info) object uploadid
$$ CL.consume
void $ sendS3Handle' h $
S3.postCompleteMultipartUpload (bucket info) object uploadid $
zip [1..] (map T.pack etags)
void $ sendS3Handle' h $ S3.postCompleteMultipartUpload
(bucket info) object uploadid (zip [1..] etags)
#else
warning $ "Cannot do multipart upload (partsize " ++ show sz ++ "); built with too old a version of the aws library."
singlepartupload k f p

8
doc/bugs/S3_upload_not_using_multipart.mdwn
View file

@ -52,3 +52,11 @@ Please provide any additional information below.
upgrade supported from repository versions: 0 1 2
[[!tag confirmed]]
> [[fixed|done]] This is now supported, when git-annex is built with a new
> enough version of the aws library. You need to configure the remote to
> use an appropriate value for multipart, eg:
>
> git annex enableremote cloud multipart=1GiB
>
> --[[Joey]]