2011-10-06 19:23:26 +00:00
|
|
|
{- git-annex uuid-based logs
|
|
|
|
-
|
2014-07-24 20:23:36 +00:00
|
|
|
- This is used to store information about UUIDs in a way that can
|
2011-10-06 19:23:26 +00:00
|
|
|
- be union merged.
|
|
|
|
-
|
2019-02-21 17:43:21 +00:00
|
|
|
- The old format looks like: "UUID[ INFO[ timestamp=foo]]"
|
2011-10-06 19:23:26 +00:00
|
|
|
- The timestamp is last for backwards compatability reasons,
|
2019-02-21 17:43:21 +00:00
|
|
|
- and may not be present on very old log lines.
|
add remote state logs
This allows a remote to store a piece of arbitrary state associated with a
key. This is needed to support Tahoe, where the file-cap is calculated from
the data stored in it, and used to retrieve a key later. Glacier also would
be much improved by using this.
GETSTATE and SETSTATE are added to the external special remote protocol.
Note that the state is left as-is even when a key is removed from a remote.
It's up to the remote to decide when it wants to clear the state.
The remote state log, $KEY.log.rmt, is a UUID-based log. However,
rather than using the old UUID-based log format, I created a new variant
of that format. The new varient is more space efficient (since it lacks the
"timestamp=" hack, and easier to parse (and the parser doesn't mess with
whitespace in the value), and avoids compatability cruft in the old one.
This seemed worth cleaning up for these new files, since there could be a
lot of them, while before UUID-based logs were only used for a few log
files at the top of the git-annex branch. The transition code has also
been updated to handle these new UUID-based logs.
This commit was sponsored by Daniel Hofer.
2014-01-03 20:35:57 +00:00
|
|
|
-
|
|
|
|
- New uuid based logs instead use the form: "timestamp UUID INFO"
|
2011-10-06 19:23:26 +00:00
|
|
|
-
|
2019-01-09 17:06:37 +00:00
|
|
|
- Copyright 2011-2019 Joey Hess <id@joeyh.name>
|
2011-10-06 19:23:26 +00:00
|
|
|
-
|
2019-03-13 19:48:14 +00:00
|
|
|
- Licensed under the GNU AGPL version 3 or higher.
|
2011-10-06 19:23:26 +00:00
|
|
|
-}
|
|
|
|
|
convert old uuid-based log parsers to attoparsec
This preserves the workaround for the old bug that caused NoUUID items
to be stored in the log, prefixing log lines with " ". It's now handled
implicitly, by using takeWhile1 (/= ' ') to get the uuid.
There is a behavior change from the old parser, which split the value
into words and then recombined it. That meant that "foo bar" and "foo\tbar"
came out as "foo bar". That behavior was not documented, and seems
surprising; it meant that after a git-annex describe here "foo bar",
you wouldn't get that same string back out when git-annex displayed repo
descriptions.
Otoh, some other parsers relied on the old behavior, and the attoparsec
rewrites had to deal with the issue themselves...
For group.log, there are some edge cases around the user providing a
group name with a leading or trailing space. The old parser would ignore
such excess whitespace. The new parser does too, because the alternative
is to refuse to parse something like " group1 group2 " due to excess
whitespace, which would be even more confusing behavior.
The only git-annex branch log file that is not converted to attoparsec
and bytestring-builder now is transitions.log.
2019-01-10 18:39:36 +00:00
|
|
|
{-# LANGUAGE OverloadedStrings, TupleSections #-}
|
2019-01-09 18:00:35 +00:00
|
|
|
|
2011-10-15 20:21:08 +00:00
|
|
|
module Logs.UUIDBased (
|
2011-10-06 19:23:26 +00:00
|
|
|
Log,
|
|
|
|
LogEntry(..),
|
2017-08-14 17:55:38 +00:00
|
|
|
VectorClock,
|
|
|
|
currentVectorClock,
|
2019-02-21 17:43:21 +00:00
|
|
|
parseLogOld,
|
add remote state logs
This allows a remote to store a piece of arbitrary state associated with a
key. This is needed to support Tahoe, where the file-cap is calculated from
the data stored in it, and used to retrieve a key later. Glacier also would
be much improved by using this.
GETSTATE and SETSTATE are added to the external special remote protocol.
Note that the state is left as-is even when a key is removed from a remote.
It's up to the remote to decide when it wants to clear the state.
The remote state log, $KEY.log.rmt, is a UUID-based log. However,
rather than using the old UUID-based log format, I created a new variant
of that format. The new varient is more space efficient (since it lacks the
"timestamp=" hack, and easier to parse (and the parser doesn't mess with
whitespace in the value), and avoids compatability cruft in the old one.
This seemed worth cleaning up for these new files, since there could be a
lot of them, while before UUID-based logs were only used for a few log
files at the top of the git-annex branch. The transition code has also
been updated to handle these new UUID-based logs.
This commit was sponsored by Daniel Hofer.
2014-01-03 20:35:57 +00:00
|
|
|
parseLogNew,
|
2019-02-21 17:43:21 +00:00
|
|
|
parseLogOldWithUUID,
|
|
|
|
buildLogOld,
|
2019-01-09 17:06:37 +00:00
|
|
|
buildLogNew,
|
2011-10-06 19:23:26 +00:00
|
|
|
changeLog,
|
|
|
|
addLog,
|
|
|
|
simpleMap,
|
|
|
|
) where
|
|
|
|
|
|
|
|
import qualified Data.Map as M
|
|
|
|
|
|
|
|
import Common
|
|
|
|
import Types.UUID
|
2017-08-14 17:55:38 +00:00
|
|
|
import Annex.VectorClock
|
2014-03-15 17:44:31 +00:00
|
|
|
import Logs.MapLog
|
2016-05-27 15:45:13 +00:00
|
|
|
import Logs.Line
|
2011-10-06 19:23:26 +00:00
|
|
|
|
convert old uuid-based log parsers to attoparsec
This preserves the workaround for the old bug that caused NoUUID items
to be stored in the log, prefixing log lines with " ". It's now handled
implicitly, by using takeWhile1 (/= ' ') to get the uuid.
There is a behavior change from the old parser, which split the value
into words and then recombined it. That meant that "foo bar" and "foo\tbar"
came out as "foo bar". That behavior was not documented, and seems
surprising; it meant that after a git-annex describe here "foo bar",
you wouldn't get that same string back out when git-annex displayed repo
descriptions.
Otoh, some other parsers relied on the old behavior, and the attoparsec
rewrites had to deal with the issue themselves...
For group.log, there are some edge cases around the user providing a
group name with a leading or trailing space. The old parser would ignore
such excess whitespace. The new parser does too, because the alternative
is to refuse to parse something like " group1 group2 " due to excess
whitespace, which would be even more confusing behavior.
The only git-annex branch log file that is not converted to attoparsec
and bytestring-builder now is transitions.log.
2019-01-10 18:39:36 +00:00
|
|
|
import qualified Data.ByteString as S
|
2019-01-10 17:23:42 +00:00
|
|
|
import qualified Data.ByteString.Lazy as L
|
2021-03-24 16:11:50 +00:00
|
|
|
import qualified Data.Attoparsec.ByteString as A
|
|
|
|
import qualified Data.Attoparsec.ByteString.Lazy as AL
|
convert old uuid-based log parsers to attoparsec
This preserves the workaround for the old bug that caused NoUUID items
to be stored in the log, prefixing log lines with " ". It's now handled
implicitly, by using takeWhile1 (/= ' ') to get the uuid.
There is a behavior change from the old parser, which split the value
into words and then recombined it. That meant that "foo bar" and "foo\tbar"
came out as "foo bar". That behavior was not documented, and seems
surprising; it meant that after a git-annex describe here "foo bar",
you wouldn't get that same string back out when git-annex displayed repo
descriptions.
Otoh, some other parsers relied on the old behavior, and the attoparsec
rewrites had to deal with the issue themselves...
For group.log, there are some edge cases around the user providing a
group name with a leading or trailing space. The old parser would ignore
such excess whitespace. The new parser does too, because the alternative
is to refuse to parse something like " group1 group2 " due to excess
whitespace, which would be even more confusing behavior.
The only git-annex branch log file that is not converted to attoparsec
and bytestring-builder now is transitions.log.
2019-01-10 18:39:36 +00:00
|
|
|
import qualified Data.Attoparsec.ByteString.Char8 as A8
|
2019-01-09 17:06:37 +00:00
|
|
|
import Data.ByteString.Builder
|
convert old uuid-based log parsers to attoparsec
This preserves the workaround for the old bug that caused NoUUID items
to be stored in the log, prefixing log lines with " ". It's now handled
implicitly, by using takeWhile1 (/= ' ') to get the uuid.
There is a behavior change from the old parser, which split the value
into words and then recombined it. That meant that "foo bar" and "foo\tbar"
came out as "foo bar". That behavior was not documented, and seems
surprising; it meant that after a git-annex describe here "foo bar",
you wouldn't get that same string back out when git-annex displayed repo
descriptions.
Otoh, some other parsers relied on the old behavior, and the attoparsec
rewrites had to deal with the issue themselves...
For group.log, there are some edge cases around the user providing a
group name with a leading or trailing space. The old parser would ignore
such excess whitespace. The new parser does too, because the alternative
is to refuse to parse something like " group1 group2 " due to excess
whitespace, which would be even more confusing behavior.
The only git-annex branch log file that is not converted to attoparsec
and bytestring-builder now is transitions.log.
2019-01-10 18:39:36 +00:00
|
|
|
import qualified Data.DList as D
|
2019-01-09 17:06:37 +00:00
|
|
|
|
2014-03-15 17:44:31 +00:00
|
|
|
type Log v = MapLog UUID v
|
2011-10-06 19:23:26 +00:00
|
|
|
|
2019-02-21 17:43:21 +00:00
|
|
|
buildLogOld :: (v -> Builder) -> Log v -> Builder
|
|
|
|
buildLogOld builder = mconcat . map genline . M.toList
|
2012-11-11 04:51:07 +00:00
|
|
|
where
|
2019-01-09 18:00:35 +00:00
|
|
|
genline (u, LogEntry c@(VectorClock {}) v) =
|
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
|
|
|
buildUUID u <> sp <> builder v <> sp
|
|
|
|
<> byteString "timestamp="
|
|
|
|
<> buildVectorClock c
|
|
|
|
<> nl
|
2019-01-09 18:00:35 +00:00
|
|
|
genline (u, LogEntry Unknown v) =
|
|
|
|
buildUUID u <> sp <> builder v <> nl
|
|
|
|
sp = charUtf8 ' '
|
|
|
|
nl = charUtf8 '\n'
|
2011-10-06 19:23:26 +00:00
|
|
|
|
2019-02-21 17:43:21 +00:00
|
|
|
parseLogOld :: A.Parser a -> L.ByteString -> Log a
|
|
|
|
parseLogOld = parseLogOldWithUUID . const
|
2012-10-10 17:52:24 +00:00
|
|
|
|
2019-02-21 17:43:21 +00:00
|
|
|
parseLogOldWithUUID :: (UUID -> A.Parser a) -> L.ByteString -> Log a
|
2021-03-24 16:11:50 +00:00
|
|
|
parseLogOldWithUUID parser = fromMaybe M.empty . AL.maybeResult
|
|
|
|
. AL.parse (logParserOld parser)
|
convert old uuid-based log parsers to attoparsec
This preserves the workaround for the old bug that caused NoUUID items
to be stored in the log, prefixing log lines with " ". It's now handled
implicitly, by using takeWhile1 (/= ' ') to get the uuid.
There is a behavior change from the old parser, which split the value
into words and then recombined it. That meant that "foo bar" and "foo\tbar"
came out as "foo bar". That behavior was not documented, and seems
surprising; it meant that after a git-annex describe here "foo bar",
you wouldn't get that same string back out when git-annex displayed repo
descriptions.
Otoh, some other parsers relied on the old behavior, and the attoparsec
rewrites had to deal with the issue themselves...
For group.log, there are some edge cases around the user providing a
group name with a leading or trailing space. The old parser would ignore
such excess whitespace. The new parser does too, because the alternative
is to refuse to parse something like " group1 group2 " due to excess
whitespace, which would be even more confusing behavior.
The only git-annex branch log file that is not converted to attoparsec
and bytestring-builder now is transitions.log.
2019-01-10 18:39:36 +00:00
|
|
|
|
2019-02-21 17:43:21 +00:00
|
|
|
logParserOld :: (UUID -> A.Parser a) -> A.Parser (Log a)
|
|
|
|
logParserOld parser = M.fromListWith best <$> parseLogLines go
|
2012-11-11 04:51:07 +00:00
|
|
|
where
|
convert old uuid-based log parsers to attoparsec
This preserves the workaround for the old bug that caused NoUUID items
to be stored in the log, prefixing log lines with " ". It's now handled
implicitly, by using takeWhile1 (/= ' ') to get the uuid.
There is a behavior change from the old parser, which split the value
into words and then recombined it. That meant that "foo bar" and "foo\tbar"
came out as "foo bar". That behavior was not documented, and seems
surprising; it meant that after a git-annex describe here "foo bar",
you wouldn't get that same string back out when git-annex displayed repo
descriptions.
Otoh, some other parsers relied on the old behavior, and the attoparsec
rewrites had to deal with the issue themselves...
For group.log, there are some edge cases around the user providing a
group name with a leading or trailing space. The old parser would ignore
such excess whitespace. The new parser does too, because the alternative
is to refuse to parse something like " group1 group2 " due to excess
whitespace, which would be even more confusing behavior.
The only git-annex branch log file that is not converted to attoparsec
and bytestring-builder now is transitions.log.
2019-01-10 18:39:36 +00:00
|
|
|
go = do
|
|
|
|
u <- toUUID <$> A8.takeWhile1 (/= ' ')
|
|
|
|
(dl, ts) <- accumval D.empty
|
|
|
|
v <- either fail return $ A.parseOnly (parser u <* A.endOfInput)
|
|
|
|
(S.intercalate " " $ D.toList dl)
|
|
|
|
return (u, LogEntry ts v)
|
|
|
|
accumval dl =
|
|
|
|
((dl,) <$> parsetimestamp)
|
|
|
|
<|> (A8.char ' ' *> (A8.takeWhile (/= ' ')) >>= accumval . D.snoc dl)
|
|
|
|
parsetimestamp =
|
|
|
|
(A8.string " timestamp=" *> vectorClockParser <* A.endOfInput)
|
|
|
|
<|> (const Unknown <$> A.endOfInput)
|
2011-10-06 19:23:26 +00:00
|
|
|
|
2019-01-09 17:06:37 +00:00
|
|
|
buildLogNew :: (v -> Builder) -> Log v -> Builder
|
2019-01-09 18:00:35 +00:00
|
|
|
buildLogNew = buildMapLog buildUUID
|
2011-10-06 19:23:26 +00:00
|
|
|
|
2019-01-10 17:23:42 +00:00
|
|
|
parseLogNew :: A.Parser v -> L.ByteString -> Log v
|
|
|
|
parseLogNew = parseMapLog (toUUID <$> A.takeByteString)
|
2011-10-06 19:23:26 +00:00
|
|
|
|
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
|
|
|
changeLog :: CandidateVectorClock -> UUID -> v -> Log v -> Log v
|
2014-03-15 17:44:31 +00:00
|
|
|
changeLog = changeMapLog
|
2011-10-06 19:23:26 +00:00
|
|
|
|
2014-03-15 17:44:31 +00:00
|
|
|
addLog :: UUID -> LogEntry v -> Log v -> Log v
|
|
|
|
addLog = addMapLog
|