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

comment

Browse source
This commit is contained in:
Joey Hess 2015-07-02 12:53:47 -04:00
parent 833cf5fff9
commit af3d4dc764
1 changed files with 21 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files

21
doc/todo/S3_fsck_support/comment_3_2b96f61188b2ada729a284e95fc89dab._comment Normal file
View file

@ -0,0 +1,21 @@
[[!comment format=mdwn
username="joey"
subject="""comment 3"""
date="2015-07-02T16:43:03Z"
content="""
Not knowing how S3's backend is implemented, I have to assume that, in
order to get the advertised number of 9's of reliability, it involves some
replication of data, as well as some method to detect if a bit has flipped
or a drive has died, and recover.
This is requesting that git-annex ask S3 for a md5sum, and compare it
against a md5sum that it, presumably, keeps track of locally. If the two
are different, git-annex could tell that S3 has lost data. But, git-annex is
in a much worse position to tell if S3 has lost data then S3 itself is.
It seems very unlikely that this extra checking would ever detect a problem
that S3 didn't itself detect and fix (or in the 0.00001% case,
fail to fix and delete the lost file?)
Bit flips during transfer seem more likely than that. `poContentMD5` could
help guard against those.
"""]]