9 lines
1.3 KiB
Text
9 lines
1.3 KiB
Text
[[!comment format=mdwn
|
|
username="Ilya_Shlyakhter"
|
|
avatar="http://cdn.libravatar.org/avatar/1647044369aa7747829c38b9dcc84df0"
|
|
subject="comment 4"
|
|
date="2018-10-08T21:43:17Z"
|
|
content="""
|
|
@joey thanks. But, besides export.log, the S3 remote also keeps some (undocumented?) internal state, and there's not way to update that state to record the fact that git-annex can GET a given key by downloading s3://mybucket/myobject ? Also, I feel uneasy directly manipulating git-annex internal files. Can you think of any plumbing commands, that could be added to support this use case?
|
|
The use case is, I submit a batch job that takes as input some s3:// objects, writes outputs to other s3:// objects, and returns pointers to these new s3:// objects. I want to register these new objects in git-annex, initially without downloading them, but be able to git-annex-get these objects, drop them from the S3 remote, but later be able to put them back under their original s3:// URIs. The latter ability is needed because (1) many workflows expect filenames to be in a particular form, e.g. mysamplename.pN.bam to represent mysample processed with parameter p=N; and (2) some workflow engines can reuse past results if a step is re-run with the same inputs, but they need the results to be at the same s3:// URI as when the step was first run.
|
|
"""]]
|