update chunk log after speculated chunks are verified to be present

Only done in checkPresentChunks, although retrieveChunks could also do
it. Does not seem necessary though, because git-annex never retrives
content without first checking if it's present AFAICR. And really this will
only be needed when using fsck. Puttting it here, rather than in fsck
avoids breaking an abstraction boundary, and is nice and inexpensive.
This commit is contained in:
Joey Hess 2020-10-22 13:37:09 -04:00
parent dad4be97c2
commit b62e004c2c
No known key found for this signature in database
GPG key ID: DB12DB0FF05F8F38
2 changed files with 38 additions and 14 deletions

View file

@ -206,7 +206,7 @@ seekResume h encryptor chunkkeys checker = do
-} -}
removeChunks :: Remover -> UUID -> ChunkConfig -> EncKey -> Key -> Annex () removeChunks :: Remover -> UUID -> ChunkConfig -> EncKey -> Key -> Annex ()
removeChunks remover u chunkconfig encryptor k = do removeChunks remover u chunkconfig encryptor k = do
ls <- chunkKeys u chunkconfig k ls <- map chunkKeyList <$> chunkKeys u chunkconfig k
mapM_ (remover . encryptor) (concat ls) mapM_ (remover . encryptor) (concat ls)
let chunksizes = catMaybes $ map (fromKey keyChunkSize <=< headMaybe) ls let chunksizes = catMaybes $ map (fromKey keyChunkSize <=< headMaybe) ls
forM_ chunksizes $ chunksRemoved u k . FixedSizeChunks . fromIntegral forM_ chunksizes $ chunksRemoved u k . FixedSizeChunks . fromIntegral
@ -245,7 +245,8 @@ retrieveChunks retriever u chunkconfig encryptor basek dest basep sink
(\e -> go (Just e) =<< chunkKeysOnly u chunkconfig basek) (\e -> go (Just e) =<< chunkKeysOnly u chunkconfig basek)
| otherwise = go Nothing =<< chunkKeys u chunkconfig basek | otherwise = go Nothing =<< chunkKeys u chunkconfig basek
where where
go pe ls = do go pe cks = do
let ls = map chunkKeyList cks
currsize <- liftIO $ catchMaybeIO $ getFileSize dest currsize <- liftIO $ catchMaybeIO $ getFileSize dest
let ls' = maybe ls (setupResume ls) currsize let ls' = maybe ls (setupResume ls) currsize
if any null ls' if any null ls'
@ -359,14 +360,18 @@ checkPresentChunks checker u chunkconfig encryptor basek
where where
checklists Nothing [] = return False checklists Nothing [] = return False
checklists (Just deferrederror) [] = throwM deferrederror checklists (Just deferrederror) [] = throwM deferrederror
checklists d (l:ls) checklists d (ck:cks)
| not (null l) = do | not (null l) = do
v <- checkchunks l v <- checkchunks l
case v of case v of
Left e -> checklists (Just e) ls Left e -> checklists (Just e) cks
Right True -> return True Right True -> do
Right False -> checklists Nothing ls ensureChunksAreLogged u basek ck
| otherwise = checklists d ls return True
Right False -> checklists Nothing cks
| otherwise = checklists d cks
where
l = chunkKeyList ck
checkchunks :: [Key] -> Annex (Either SomeException Bool) checkchunks :: [Key] -> Annex (Either SomeException Bool)
checkchunks [] = return (Right True) checkchunks [] = return (Right True)
@ -379,6 +384,14 @@ checkPresentChunks checker u chunkconfig encryptor basek
check = tryNonAsync . checker . encryptor check = tryNonAsync . checker . encryptor
data ChunkKeys
= ChunkKeys [Key]
| SpeculativeChunkKeys (ChunkMethod, ChunkCount) [Key]
chunkKeyList :: ChunkKeys -> [Key]
chunkKeyList (ChunkKeys l) = l
chunkKeyList (SpeculativeChunkKeys _ l) = l
{- A key can be stored in a remote unchunked, or as a list of chunked keys. {- A key can be stored in a remote unchunked, or as a list of chunked keys.
- This can be the case whether or not the remote is currently configured - This can be the case whether or not the remote is currently configured
- to use chunking. - to use chunking.
@ -395,22 +408,22 @@ checkPresentChunks checker u chunkconfig encryptor basek
- recover from data loss, where the chunk log didn't make it out, - recover from data loss, where the chunk log didn't make it out,
- though only as long as the ChunkConfig is unchanged. - though only as long as the ChunkConfig is unchanged.
-} -}
chunkKeys :: UUID -> ChunkConfig -> Key -> Annex [[Key]] chunkKeys :: UUID -> ChunkConfig -> Key -> Annex [ChunkKeys]
chunkKeys = chunkKeys' False chunkKeys = chunkKeys' False
{- Same as chunkKeys, but excluding the unchunked key. -} {- Same as chunkKeys, but excluding the unchunked key. -}
chunkKeysOnly :: UUID -> ChunkConfig -> Key -> Annex [[Key]] chunkKeysOnly :: UUID -> ChunkConfig -> Key -> Annex [ChunkKeys]
chunkKeysOnly = chunkKeys' True chunkKeysOnly = chunkKeys' True
chunkKeys' :: Bool -> UUID -> ChunkConfig -> Key -> Annex [[Key]] chunkKeys' :: Bool -> UUID -> ChunkConfig -> Key -> Annex [ChunkKeys]
chunkKeys' onlychunks u chunkconfig k = do chunkKeys' onlychunks u chunkconfig k = do
recorded <- getCurrentChunks u k recorded <- getCurrentChunks u k
let recordedl = map (toChunkList k) recorded let recordedl = map (ChunkKeys . toChunkList k) recorded
return $ addspeculative recorded $ if onlychunks return $ addspeculative recorded $ if onlychunks
then recordedl then recordedl
else if noChunks chunkconfig else if noChunks chunkconfig
then [k] : recordedl then ChunkKeys [k] : recordedl
else recordedl ++ [[k]] else recordedl ++ [ChunkKeys [k]]
where where
addspeculative recorded l = case chunkconfig of addspeculative recorded l = case chunkconfig of
NoChunks -> l NoChunks -> l
@ -422,10 +435,19 @@ chunkKeys' onlychunks u chunkconfig k = do
v = (FixedSizeChunks chunksz, chunkcount) v = (FixedSizeChunks chunksz, chunkcount)
in if v `elem` recorded in if v `elem` recorded
then l then l
else l ++ [toChunkList k v] else l ++ [SpeculativeChunkKeys v (toChunkList k v)]
LegacyChunks _ -> l LegacyChunks _ -> l
toChunkList :: Key -> (ChunkMethod, ChunkCount) -> [Key] toChunkList :: Key -> (ChunkMethod, ChunkCount) -> [Key]
toChunkList k (FixedSizeChunks chunksize, chunkcount) = toChunkList k (FixedSizeChunks chunksize, chunkcount) =
takeChunkKeyStream chunkcount $ chunkKeyStream k chunksize takeChunkKeyStream chunkcount $ chunkKeyStream k chunksize
toChunkList _ (UnknownChunks _, _) = [] toChunkList _ (UnknownChunks _, _) = []
{- When chunkKeys provided a speculative chunk list, and that has been
- verified to be present, use this to log it in the chunk log. This way,
- a later change to the chunk size of the remote won't prevent accessing
- the chunks. -}
ensureChunksAreLogged :: UUID -> Key -> ChunkKeys -> Annex ()
ensureChunksAreLogged u k (SpeculativeChunkKeys (chunkmethod, chunkcount) _) =
chunksStored u k chunkmethod chunkcount
ensureChunksAreLogged _ _ (ChunkKeys _) = return ()

View file

@ -24,3 +24,5 @@ but `annex fsck --key KEY --from remote --fast`, since it doesn't have an exact
[[!meta author=yoh]] [[!meta author=yoh]]
[[!tag projects/dandi]] [[!tag projects/dandi]]
> [[done]] --[[Joey]]