git-annex/Logs/Location.hs

256 lines
7.7 KiB
Haskell
Raw Normal View History

{-# LANGUAGE BangPatterns #-}
2010-10-09 23:22:40 +00:00
{- git-annex location log
-
2011-04-02 19:50:51 +00:00
- git-annex keeps track of which repositories have the contents of annexed
- files.
2010-10-09 23:22:40 +00:00
-
2010-10-12 22:06:34 +00:00
- Repositories record their UUID and the date when they --get or --drop
2010-10-13 00:04:36 +00:00
- a value.
2010-10-10 16:31:14 +00:00
-
- Copyright 2010-2024 Joey Hess <id@joeyh.name>
2010-10-27 20:53:54 +00:00
-
- Licensed under the GNU AGPL version 3 or higher.
2010-10-09 23:22:40 +00:00
-}
{-# LANGUAGE BangPatterns #-}
2011-10-15 20:21:08 +00:00
module Logs.Location (
2010-10-12 22:25:41 +00:00
LogStatus(..),
2012-12-12 23:20:38 +00:00
logStatus,
logStatusAfter,
logChange,
loggedLocations,
loggedPreviousLocations,
loggedLocationsHistorical,
2016-07-17 19:15:08 +00:00
loggedLocationsRef,
parseLoggedLocations,
isKnownKey,
checkDead,
setDead,
Unchecked,
finishCheck,
2011-07-01 21:23:01 +00:00
loggedKeys,
loggedKeysFor,
loggedKeysFor',
overLocationLogs,
overLocationLogs',
2010-10-11 21:52:46 +00:00
) where
2010-10-09 23:22:40 +00:00
import Annex.Common
2011-10-04 04:40:47 +00:00
import qualified Annex.Branch
import Logs
2011-10-15 20:21:08 +00:00
import Logs.Presence
remove dead nodes when loading the cluster log This is to avoid inserting a cluster uuid into the location log when only dead nodes in the cluster contain the content of a key. One reason why this is necessary is Remote.keyLocations, which excludes dead repositories from the list. But there are probably many more. Implementing this was challenging, because Logs.Location importing Logs.Cluster which imports Logs.Trust which imports Remote.List resulted in an import cycle through several other modules. Resorted to making Logs.Location not import Logs.Cluster, and instead it assumes that Annex.clusters gets populated when necessary before it's called. That's done in Annex.Startup, which is run by the git-annex command (but not other commands) at early startup in initialized repos. Or, is run after initialization. Note that is Remote.Git, it is unable to import Annex.Startup, because Remote.Git importing Logs.Cluster leads the the same import cycle. So ensureInitialized is not passed annexStartup in there. Other commands, like git-annex-shell currently don't run annexStartup either. So there are cases where Logs.Location will not see clusters. So it won't add any cluster UUIDs when loading the log. That's ok, the only reason to do that is to make display of where objects are located include clusters, and to make commands like git-annex get --from treat keys as being located in a cluster. git-annex-shell certainly does not do anything like that, and I'm pretty sure Remote.Git (and callers to Remote.Git.onLocalRepo) don't either.
2024-06-16 18:35:07 +00:00
import Types.Cluster
2012-12-12 23:20:38 +00:00
import Annex.UUID
2016-07-17 19:15:08 +00:00
import Annex.CatFile
import Annex.VectorClock
2016-07-17 19:15:08 +00:00
import Git.Types (RefDate, Ref)
import qualified Annex
2012-12-12 23:20:38 +00:00
import Data.Time.Clock
import qualified Data.ByteString.Lazy as L
import qualified Data.Map as M
import qualified Data.Set as S
2012-12-12 23:20:38 +00:00
{- Log a change in the presence of a key's value in current repository. -}
logStatus :: Key -> LogStatus -> Annex ()
2013-09-05 01:37:13 +00:00
logStatus key s = do
2012-12-12 23:20:38 +00:00
u <- getUUID
2013-09-05 01:37:13 +00:00
logChange key u s
2010-10-09 23:22:40 +00:00
{- Run an action that gets the content of a key, and update the log
- when it succeeds. -}
logStatusAfter :: Key -> Annex Bool -> Annex Bool
logStatusAfter key a = ifM a
( do
logStatus key InfoPresent
return True
, return False
)
{- Log a change in the presence of a key's value in a repository.
-
- Cluster UUIDs are not logged. Instead, when a node of a cluster is
- logged to contain a key, loading the log will include the cluster's
- UUID.
-}
2011-11-09 05:15:51 +00:00
logChange :: Key -> UUID -> LogStatus -> Annex ()
logChange key u@(UUID _) s
| isClusterUUID u = noop
| otherwise = do
config <- Annex.getGitConfig
maybeAddLog
(Annex.Branch.RegardingUUID [u])
(locationLogFile config key)
s
(LogInfo (fromUUID u))
deal better with clock skew situations, using vector clocks * Deal with clock skew, both forwards and backwards, when logging information to the git-annex branch. * GIT_ANNEX_VECTOR_CLOCK can now be set to a fixed value (eg 1) rather than needing to be advanced each time a new change is made. * Misuse of GIT_ANNEX_VECTOR_CLOCK will no longer confuse git-annex. When changing a file in the git-annex branch, the vector clock to use is now determined by first looking at the current time (or GIT_ANNEX_VECTOR_CLOCK when set), and comparing it to the newest vector clock already in use in that file. If a newer time stamp was already in use, advance it forward by a second instead. When the clock is set to a time in the past, this avoids logging with an old timestamp, which would risk that log line later being ignored in favor of "newer" line that is really not newer. When a log entry has been made with a clock that was set far ahead in the future, this avoids newer information being logged with an older timestamp and so being ignored in favor of that future-timestamped information. Once all clocks get fixed, this will result in the vector clocks being incremented, until finally enough time has passed that time gets back ahead of the vector clock value, and then it will return to usual operation. (This latter situation is not ideal, but it seems the best that can be done. The issue with it is, since all writers will be incrementing the last vector clock they saw, there's no way to tell when one writer made a write significantly later in time than another, so the earlier write might arbitrarily be picked when merging. This problem is why git-annex uses timestamps in the first place, rather than pure vector clocks.) Advancing forward by 1 second is somewhat arbitrary. setDead advances a timestamp by just 1 picosecond, and the vector clock could too. But then it would interfere with setDead, which wants to be overrulled by any change. So it could use 2 picoseconds or something, but that seems weird. It could just as well advance it forward by a minute or whatever, but then it would be harder for real time to catch up with the vector clock when forward clock slew had happened. A complication is that many log files contain several different peices of information, and it may be best to only use vector clocks for the same peice of information. For example, a key's location log file contains InfoPresent/InfoMissing for each UUID, and it only looks at the vector clocks for the UUID that is being changed, and not other UUIDs. Although exactly where the dividing line is can be hard to determine. Consider metadata logs, where a field "tag" can have multiple values set at different times. Should it advance forward past the last tag? Probably. What about when a different field is set, should it look at the clocks of other fields? Perhaps not, but currently it does, and this does not seems like it will cause any problems. Another one I'm not entirely sure about is the export log, which is keyed by (fromuuid, touuid). So if multiple repos are exporting to the same remote, different vector clocks can be used for that remote. It looks like that's probably ok, because it does not try to determine what order things occurred when there was an export conflict. Sponsored-by: Jochen Bartl on Patreon
2021-08-03 20:45:20 +00:00
logChange _ NoUUID _ = noop
2010-10-12 22:25:41 +00:00
{- Returns a list of repository UUIDs that, according to the log, have
- the value of a key. -}
loggedLocations :: Key -> Annex [UUID]
loggedLocations = getLoggedLocations presentLogInfo
{- Returns a list of repository UUIDs that the location log indicates
- used to have the vale of a key, but no longer do.
-}
loggedPreviousLocations :: Key -> Annex [UUID]
loggedPreviousLocations = getLoggedLocations notPresentLogInfo
{- Gets the location log on a particular date. -}
loggedLocationsHistorical :: RefDate -> Key -> Annex [UUID]
loggedLocationsHistorical = getLoggedLocations . historicalLogInfo
2016-07-17 19:15:08 +00:00
{- Gets the locations contained in a git ref. -}
loggedLocationsRef :: Ref -> Annex [UUID]
loggedLocationsRef ref = map (toUUID . fromLogInfo) . getLog <$> catObject ref
2016-07-17 19:15:08 +00:00
{- Parses the content of a log file and gets the locations in it. -}
parseLoggedLocations :: Clusters -> L.ByteString -> [UUID]
parseLoggedLocations clusters l = addClusterUUIDs clusters $
map (toUUID . fromLogInfo . info)
(filterPresent (parseLog l))
getLoggedLocations :: (RawFilePath -> Annex [LogInfo]) -> Key -> Annex [UUID]
getLoggedLocations getter key = do
config <- Annex.getGitConfig
locs <- map (toUUID . fromLogInfo) <$> getter (locationLogFile config key)
clusters <- getClusters
return $ addClusterUUIDs clusters locs
-- Add UUIDs of any clusters whose nodes are in the list.
addClusterUUIDs :: Clusters -> [UUID] -> [UUID]
addClusterUUIDs clusters locs
| M.null clustermap = locs
-- ^ optimisation for common case of no clusters
| otherwise = clusterlocs ++ locs
where
clustermap = clusterNodeUUIDs clusters
clusterlocs = map fromClusterUUID $ S.toList $
S.unions $ mapMaybe findclusters locs
findclusters u = M.lookup (ClusterNodeUUID u) clustermap
{- Is there a location log for the key? True even for keys with no
- remaining locations. -}
isKnownKey :: Key -> Annex Bool
isKnownKey key = do
config <- Annex.getGitConfig
not . null <$> readLog (locationLogFile config key)
{- For a key to be dead, all locations that have location status for the key
- must have InfoDead set. -}
checkDead :: Key -> Annex Bool
checkDead key = do
config <- Annex.getGitConfig
ls <- compactLog <$> readLog (locationLogFile config key)
return $! all (\l -> status l == InfoDead) ls
{- Updates the log to say that a key is dead.
-
- Changes all logged lines for the key, in any location, that are
- currently InfoMissing, to be InfoDead.
deal better with clock skew situations, using vector clocks * Deal with clock skew, both forwards and backwards, when logging information to the git-annex branch. * GIT_ANNEX_VECTOR_CLOCK can now be set to a fixed value (eg 1) rather than needing to be advanced each time a new change is made. * Misuse of GIT_ANNEX_VECTOR_CLOCK will no longer confuse git-annex. When changing a file in the git-annex branch, the vector clock to use is now determined by first looking at the current time (or GIT_ANNEX_VECTOR_CLOCK when set), and comparing it to the newest vector clock already in use in that file. If a newer time stamp was already in use, advance it forward by a second instead. When the clock is set to a time in the past, this avoids logging with an old timestamp, which would risk that log line later being ignored in favor of "newer" line that is really not newer. When a log entry has been made with a clock that was set far ahead in the future, this avoids newer information being logged with an older timestamp and so being ignored in favor of that future-timestamped information. Once all clocks get fixed, this will result in the vector clocks being incremented, until finally enough time has passed that time gets back ahead of the vector clock value, and then it will return to usual operation. (This latter situation is not ideal, but it seems the best that can be done. The issue with it is, since all writers will be incrementing the last vector clock they saw, there's no way to tell when one writer made a write significantly later in time than another, so the earlier write might arbitrarily be picked when merging. This problem is why git-annex uses timestamps in the first place, rather than pure vector clocks.) Advancing forward by 1 second is somewhat arbitrary. setDead advances a timestamp by just 1 picosecond, and the vector clock could too. But then it would interfere with setDead, which wants to be overrulled by any change. So it could use 2 picoseconds or something, but that seems weird. It could just as well advance it forward by a minute or whatever, but then it would be harder for real time to catch up with the vector clock when forward clock slew had happened. A complication is that many log files contain several different peices of information, and it may be best to only use vector clocks for the same peice of information. For example, a key's location log file contains InfoPresent/InfoMissing for each UUID, and it only looks at the vector clocks for the UUID that is being changed, and not other UUIDs. Although exactly where the dividing line is can be hard to determine. Consider metadata logs, where a field "tag" can have multiple values set at different times. Should it advance forward past the last tag? Probably. What about when a different field is set, should it look at the clocks of other fields? Perhaps not, but currently it does, and this does not seems like it will cause any problems. Another one I'm not entirely sure about is the export log, which is keyed by (fromuuid, touuid). So if multiple repos are exporting to the same remote, different vector clocks can be used for that remote. It looks like that's probably ok, because it does not try to determine what order things occurred when there was an export conflict. Sponsored-by: Jochen Bartl on Patreon
2021-08-03 20:45:20 +00:00
-
- The vector clock in the log is updated minimally, so that any
- other location log changes are guaranteed to overrule this.
-}
setDead :: Key -> Annex ()
setDead key = do
config <- Annex.getGitConfig
let logfile = locationLogFile config key
ls <- compactLog <$> readLog logfile
mapM_ (go logfile) (filter (\l -> status l == InfoMissing) ls)
where
start implementing hidden git-annex repositories This adds a separate journal, which does not currently get committed to an index, but is planned to be committed to .git/annex/index-private. Changes that are regarding a UUID that is private will get written to this journal, and so will not be published into the git-annex branch. All log writing should have been made to indicate the UUID it's regarding, though I've not verified this yet. Currently, no UUIDs are treated as private yet, a way to configure that is needed. The implementation is careful to not add any additional IO work when privateUUIDsKnown is False. It will skip looking at the private journal at all. So this should be free, or nearly so, unless the feature is used. When it is used, all branch reads will be about twice as expensive. It is very lucky -- or very prudent design -- that Annex.Branch.change and maybeChange are the only ways to change a file on the branch, and Annex.Branch.set is only internal use. That let Annex.Branch.get always yield any private information that has been recorded, without the risk that Annex.Branch.set might be called, with a non-private UUID, and end up leaking the private information into the git-annex branch. And, this relies on the way git-annex union merges the git-annex branch. When reading a file, there can be a public and a private version, and they are just concacenated together. That will be handled the same as if there were two diverged git-annex branches that got union merged.
2021-04-20 18:32:41 +00:00
go logfile l =
let u = toUUID (fromLogInfo (info l))
deal better with clock skew situations, using vector clocks * Deal with clock skew, both forwards and backwards, when logging information to the git-annex branch. * GIT_ANNEX_VECTOR_CLOCK can now be set to a fixed value (eg 1) rather than needing to be advanced each time a new change is made. * Misuse of GIT_ANNEX_VECTOR_CLOCK will no longer confuse git-annex. When changing a file in the git-annex branch, the vector clock to use is now determined by first looking at the current time (or GIT_ANNEX_VECTOR_CLOCK when set), and comparing it to the newest vector clock already in use in that file. If a newer time stamp was already in use, advance it forward by a second instead. When the clock is set to a time in the past, this avoids logging with an old timestamp, which would risk that log line later being ignored in favor of "newer" line that is really not newer. When a log entry has been made with a clock that was set far ahead in the future, this avoids newer information being logged with an older timestamp and so being ignored in favor of that future-timestamped information. Once all clocks get fixed, this will result in the vector clocks being incremented, until finally enough time has passed that time gets back ahead of the vector clock value, and then it will return to usual operation. (This latter situation is not ideal, but it seems the best that can be done. The issue with it is, since all writers will be incrementing the last vector clock they saw, there's no way to tell when one writer made a write significantly later in time than another, so the earlier write might arbitrarily be picked when merging. This problem is why git-annex uses timestamps in the first place, rather than pure vector clocks.) Advancing forward by 1 second is somewhat arbitrary. setDead advances a timestamp by just 1 picosecond, and the vector clock could too. But then it would interfere with setDead, which wants to be overrulled by any change. So it could use 2 picoseconds or something, but that seems weird. It could just as well advance it forward by a minute or whatever, but then it would be harder for real time to catch up with the vector clock when forward clock slew had happened. A complication is that many log files contain several different peices of information, and it may be best to only use vector clocks for the same peice of information. For example, a key's location log file contains InfoPresent/InfoMissing for each UUID, and it only looks at the vector clocks for the UUID that is being changed, and not other UUIDs. Although exactly where the dividing line is can be hard to determine. Consider metadata logs, where a field "tag" can have multiple values set at different times. Should it advance forward past the last tag? Probably. What about when a different field is set, should it look at the clocks of other fields? Perhaps not, but currently it does, and this does not seems like it will cause any problems. Another one I'm not entirely sure about is the export log, which is keyed by (fromuuid, touuid). So if multiple repos are exporting to the same remote, different vector clocks can be used for that remote. It looks like that's probably ok, because it does not try to determine what order things occurred when there was an export conflict. Sponsored-by: Jochen Bartl on Patreon
2021-08-03 20:45:20 +00:00
c = case date l of
VectorClock v -> CandidateVectorClock $
v + realToFrac (picosecondsToDiffTime 1)
Unknown -> CandidateVectorClock 0
in addLog' (Annex.Branch.RegardingUUID [u]) logfile InfoDead
(info l) c
2011-04-02 19:50:51 +00:00
data Unchecked a = Unchecked (Annex (Maybe a))
finishCheck :: Unchecked a -> Annex (Maybe a)
finishCheck (Unchecked a) = a
{- Finds all keys that have location log information.
- (There may be duplicate keys in the list.)
-
- Keys that have been marked as dead are not included.
-}
loggedKeys :: Annex (Maybe ([Unchecked Key], IO Bool))
loggedKeys = loggedKeys' (not <$$> checkDead)
loggedKeys' :: (Key -> Annex Bool) -> Annex (Maybe ([Unchecked Key], IO Bool))
loggedKeys' check = do
config <- Annex.getGitConfig
Annex.Branch.files >>= \case
Nothing -> return Nothing
Just (bfs, cleanup) -> do
let l = mapMaybe (defercheck <$$> locationLogFileKey config) bfs
return (Just (l, cleanup))
where
defercheck k = Unchecked $ ifM (check k)
( return (Just k)
, return Nothing
)
2011-07-01 21:23:01 +00:00
{- Finds all keys that have location log information indicating
- they are present in the specified repository.
-
- This does not stream well; use loggedKeysFor' for lazy streaming.
-}
loggedKeysFor :: UUID -> Annex (Maybe [Key])
loggedKeysFor u = loggedKeysFor' u >>= \case
Nothing -> return Nothing
Just (l, cleanup) -> do
l' <- catMaybes <$> mapM finishCheck l
liftIO $ void cleanup
return (Just l')
loggedKeysFor' :: UUID -> Annex (Maybe ([Unchecked Key], IO Bool))
loggedKeysFor' u = loggedKeys' isthere
2012-11-11 04:51:07 +00:00
where
isthere k = do
us <- loggedLocations k
let !there = u `elem` us
return there
{- This is much faster than loggedKeys. -}
overLocationLogs :: Bool -> v -> (Key -> [UUID] -> v -> Annex v) -> Annex (Annex.Branch.UnmergedBranches v)
overLocationLogs omitnewjournalledfiles v =
overLocationLogs' omitnewjournalledfiles v (flip const)
overLocationLogs'
:: Bool
-> v
-> (Annex (Maybe (Key, RawFilePath, Maybe L.ByteString)) -> Annex v -> Annex v)
-> (Key -> [UUID] -> v -> Annex v)
-> Annex (Annex.Branch.UnmergedBranches v)
overLocationLogs' omitnewjournalledfiles iv discarder keyaction = do
config <- Annex.getGitConfig
clusters <- getClusters
let getk = locationLogFileKey config
let go v reader = reader >>= \case
Just (k, f, content) -> discarder reader $ do
-- precache to make checkDead fast, and also to
-- make any accesses done in keyaction fast.
maybe noop (Annex.Branch.precache f) content
ifM (checkDead k)
( go v reader
, do
!v' <- keyaction k (maybe [] (parseLoggedLocations clusters) content) v
go v' reader
)
Nothing -> return v
Annex.Branch.overBranchFileContents getk omitnewjournalledfiles (go iv)
remove dead nodes when loading the cluster log This is to avoid inserting a cluster uuid into the location log when only dead nodes in the cluster contain the content of a key. One reason why this is necessary is Remote.keyLocations, which excludes dead repositories from the list. But there are probably many more. Implementing this was challenging, because Logs.Location importing Logs.Cluster which imports Logs.Trust which imports Remote.List resulted in an import cycle through several other modules. Resorted to making Logs.Location not import Logs.Cluster, and instead it assumes that Annex.clusters gets populated when necessary before it's called. That's done in Annex.Startup, which is run by the git-annex command (but not other commands) at early startup in initialized repos. Or, is run after initialization. Note that is Remote.Git, it is unable to import Annex.Startup, because Remote.Git importing Logs.Cluster leads the the same import cycle. So ensureInitialized is not passed annexStartup in there. Other commands, like git-annex-shell currently don't run annexStartup either. So there are cases where Logs.Location will not see clusters. So it won't add any cluster UUIDs when loading the log. That's ok, the only reason to do that is to make display of where objects are located include clusters, and to make commands like git-annex get --from treat keys as being located in a cluster. git-annex-shell certainly does not do anything like that, and I'm pretty sure Remote.Git (and callers to Remote.Git.onLocalRepo) don't either.
2024-06-16 18:35:07 +00:00
-- Cannot import Logs.Cluster due to a cycle.
-- Annex.clusters gets populated when starting up git-annex.
getClusters :: Annex Clusters
getClusters = maybe (pure noClusters) id =<< Annex.getState Annex.clusters