Joey Hess [Fri, 20 Sep 2024 19:39:52 +0000 (15:39 -0400)]
bugfixes
sim stabilization works now
Joey Hess [Fri, 20 Sep 2024 19:21:18 +0000 (15:21 -0400)]
sim: better step
On each step, find all the actions that could be done, and pick one of them
to do.
Should detect stability, but that is broken.
Joey Hess [Fri, 20 Sep 2024 19:11:55 +0000 (15:11 -0400)]
set simRootDirectory on restore
It's a relative directory and the cwd may be different. Or the repo
could have been moved.
Joey Hess [Fri, 20 Sep 2024 18:57:55 +0000 (14:57 -0400)]
remove sim log file
Joey Hess [Fri, 20 Sep 2024 15:59:35 +0000 (11:59 -0400)]
update
Joey Hess [Fri, 20 Sep 2024 15:26:40 +0000 (11:26 -0400)]
update
Joey Hess [Fri, 20 Sep 2024 15:05:57 +0000 (11:05 -0400)]
update
Joey Hess [Fri, 20 Sep 2024 14:34:52 +0000 (10:34 -0400)]
sim: implement addtree
Joey Hess [Tue, 17 Sep 2024 18:39:53 +0000 (14:39 -0400)]
sim: tested concurrency over actions
This demonstrates concurrent behavior that looks right. And with a
random seed, the results are deterministic.
init foo
init bar
init backup
connect foo <-> bar
connect foo <-> backup
addmulti 10 testfiles 1mb 1gb foo backup
action foo gitpull backup
wanted foo nothing
wanted bar anything
wanted backup anything
action bar gitpull foo
action foo dropunwanted while action bar getwanted foo
Joey Hess [Tue, 17 Sep 2024 18:29:49 +0000 (14:29 -0400)]
sim: fix get bug
When getting from a remote, have to check that the repo doing the
getting thinks the remote contains the key, but also that the remote
actually does. Before this bug fix, it would get from a repo that used
to have the key, but that had dropped it since the last git pull.
Joey Hess [Tue, 17 Sep 2024 17:49:50 +0000 (13:49 -0400)]
record initial seed in sim log
Unless the log starts with a command that records a seed.
Joey Hess [Tue, 17 Sep 2024 17:43:11 +0000 (13:43 -0400)]
git-annex sim log
Joey Hess [Tue, 17 Sep 2024 17:35:27 +0000 (13:35 -0400)]
sim: implement dropunwantedfrom
Joey Hess [Tue, 17 Sep 2024 15:55:57 +0000 (11:55 -0400)]
fix ActionDropUnwanted
Now tested working
Joey Hess [Tue, 17 Sep 2024 15:40:40 +0000 (11:40 -0400)]
sim: add missing parser for action sendwanted
Joey Hess [Tue, 17 Sep 2024 15:37:25 +0000 (11:37 -0400)]
sim: fix changing settings bug
Joey Hess [Tue, 17 Sep 2024 15:19:59 +0000 (11:19 -0400)]
addmulti
Joey Hess [Tue, 17 Sep 2024 14:50:25 +0000 (10:50 -0400)]
move pull/push/sync into getSimActionComponents
As well as being a more pleasing implementation than I managed
yesterday, this allows for those actions to be run concurrently in the
sim.
Joey Hess [Tue, 17 Sep 2024 13:33:55 +0000 (09:33 -0400)]
sim concurrency
Joey Hess [Tue, 17 Sep 2024 12:31:55 +0000 (08:31 -0400)]
add
Joey Hess [Mon, 16 Sep 2024 20:58:03 +0000 (16:58 -0400)]
implement ActionPull/Pull/Sync
Somewhat unsatisfying implementation, but all the alternatives were
worse.
Joey Hess [Mon, 16 Sep 2024 20:48:21 +0000 (16:48 -0400)]
sim: add action repo sync command
Joey Hess [Thu, 12 Sep 2024 20:39:44 +0000 (16:39 -0400)]
sim command fixes
Joey Hess [Thu, 12 Sep 2024 20:07:44 +0000 (16:07 -0400)]
git-annex sim command is working
Had to add Read instances to Key and NumCopies and some other similar
types. I only expect to use those in serializing a sim. Of course, this
risks that implementation changes break reading old data. For a sim,
that would not be a big problem.
Joey Hess [Thu, 12 Sep 2024 14:44:31 +0000 (10:44 -0400)]
implemented ActionDropUnwanted
Not tested yet. This emulates the same checking that is done when
dropping. Note that when dropping from a special remote it is not able
to make a locked copy.
Joey Hess [Thu, 12 Sep 2024 14:08:50 +0000 (10:08 -0400)]
convert simRepoState to use UUID as key
Joey Hess [Thu, 12 Sep 2024 01:00:51 +0000 (21:00 -0400)]
fully preserve input format of connect and disconnect commands
Just lifted the sim file as a DSL into the type level for that.
Joey Hess [Wed, 11 Sep 2024 19:59:13 +0000 (15:59 -0400)]
sim file parser and generator
The generator doesn't emit the best possible connect commands,
but it does output something valid. Eg, an input like:
connect A <-> B <-> C <-> D
becomes:
connect A <-> B <-> C
connect C <-> D
Also:
connect A -> B <- C
becomes:
connect A -> B
connect C -> B
Which could be improved.
Also disconnect commands are not prettified at all, but probably there's
no reason to.
Joey Hess [Wed, 11 Sep 2024 15:53:25 +0000 (11:53 -0400)]
started on sim file parser
Joey Hess [Wed, 11 Sep 2024 15:04:48 +0000 (11:04 -0400)]
implemented ActionSendWanted
The simulation is nearly finished. Only CommandAddTree and dropping
remain to do.
Joey Hess [Wed, 11 Sep 2024 14:32:04 +0000 (10:32 -0400)]
ActionGetWanted working
The sim is now basically working!
Joey Hess [Tue, 10 Sep 2024 19:37:55 +0000 (15:37 -0400)]
sim: implement ActionGitPush and ActionGitPull
These don't actually need to do any git actions, instead they just merge
the simLocations.
Had to make simLocations use a vector clock, because it is possible for
two simulated repositories to end up with different opinions about the
location of a key. Just like with real git-annex, whichever location
change was made most recently wins out.
The vector clock is simply advanced each time the simulation is run for
a step. Since there is no real parallelism in the sim, that's
sufficient.
Joey Hess [Mon, 9 Sep 2024 21:20:13 +0000 (17:20 -0400)]
implement randomAction
Joey Hess [Mon, 9 Sep 2024 21:04:32 +0000 (17:04 -0400)]
implement CommandStep
and record each SimCommand in simHistory, except for CommandStep where
instead the actions that are generated are recorded.
Joey Hess [Mon, 9 Sep 2024 20:06:45 +0000 (16:06 -0400)]
add action command to git-annex sim
step just picks a random action, and this allows finer control over what
happens in the sim
Joey Hess [Mon, 9 Sep 2024 18:52:24 +0000 (14:52 -0400)]
set location logs in simulated repos
Joey Hess [Mon, 9 Sep 2024 18:07:52 +0000 (14:07 -0400)]
fix removing simulated file
Joey Hess [Mon, 9 Sep 2024 15:28:30 +0000 (11:28 -0400)]
create simulated files
Joey Hess [Mon, 9 Sep 2024 15:06:42 +0000 (11:06 -0400)]
set descriptions for all simulated repos
Joey Hess [Mon, 9 Sep 2024 14:59:01 +0000 (10:59 -0400)]
simulated repository construction working
Joey Hess [Mon, 9 Sep 2024 13:35:42 +0000 (09:35 -0400)]
finish updateSimRepoState
Converted maps to use UUID as key.
Also added mincopies to the sim.
Joey Hess [Mon, 9 Sep 2024 13:32:43 +0000 (09:32 -0400)]
maxsize of 0 to disable
Joey Hess [Fri, 6 Sep 2024 18:23:29 +0000 (14:23 -0400)]
implemented cloneSimRepo
Started on updateSimRepoState
Joey Hess [Fri, 6 Sep 2024 16:53:51 +0000 (12:53 -0400)]
more work on applySimCommand
When using an existing repo, copy over all of its config into the sim.
Added CommandTrustLevel.
Start at creating a git clone for a simulated repo, but it's not done
yet.
Joey Hess [Fri, 6 Sep 2024 16:53:20 +0000 (12:53 -0400)]
rethought sim a bit
Joey Hess [Thu, 5 Sep 2024 20:22:08 +0000 (16:22 -0400)]
prevent overwriting a repo in the simulation
Joey Hess [Thu, 5 Sep 2024 19:22:41 +0000 (15:22 -0400)]
almost finished with applySimCommand
Added checks that repo names are ones that have been added to the sim.
Implemented preferred content etc setting. It does not need to parse the
expression in applySimCommand, instead that can be done when running the
sim. This keeps it pure.
But, it can't be entirely pure because of CommandAddTree. So made it
return an Annex action when necessary.
Moved makeMatcher into Annex.FileMatcher in preparation for using it,
but it's not used yet. Also moved checkPreferredContentExpression.
Joey Hess [Thu, 5 Sep 2024 14:50:04 +0000 (10:50 -0400)]
implement CommandUse in Annex.Sim
Refactored Remote to keep it pure.
Joey Hess [Wed, 4 Sep 2024 19:10:39 +0000 (15:10 -0400)]
started Annex.Sim
Have most of the sim command handler, but to keep it pure while implementing
the rest will need some refactoring.
It seems likely that running the simulation itself will not be able to be
entirely pure. Preferred content evaluation runs in Annex after all.
Note that the somewhat awkward randomWords is because the i386ancient
build depends on a version of random too old to support generating a
random ByteString on its own.
Joey Hess [Wed, 4 Sep 2024 19:03:17 +0000 (15:03 -0400)]
documentation for git-annex sim
command not implemented yet
Joey Hess [Tue, 3 Sep 2024 18:24:32 +0000 (14:24 -0400)]
update
Joey Hess [Tue, 3 Sep 2024 18:23:48 +0000 (14:23 -0400)]
sim design
Joey Hess [Tue, 3 Sep 2024 18:23:32 +0000 (14:23 -0400)]
correct comment
This is not related to v5 versus newer versions.
Joey Hess [Tue, 3 Sep 2024 18:23:14 +0000 (14:23 -0400)]
fix typo
Joey Hess [Tue, 3 Sep 2024 17:49:34 +0000 (13:49 -0400)]
treat "not present" in preferred content as invalid
Detect when a preferred content expression contains "not present", which
would lead to repeatedly getting and then dropping files, and make it never
match. This also applies to "not balanced" and "not sizebalanced".
--explain will tell the user when this happens
Note that getMatcher calls matchMrun' and does not check for unstable
negated limits. While there is no --present anyway, if there was,
it would not make sense for --not --present to complain about
instability and fail to match.
Joey Hess [Tue, 3 Sep 2024 16:38:59 +0000 (12:38 -0400)]
Fix --debug display of onlyingroup preferred content expression.
Joey Hess [Tue, 3 Sep 2024 15:52:54 +0000 (11:52 -0400)]
update
Joey Hess [Mon, 2 Sep 2024 20:32:28 +0000 (16:32 -0400)]
2 level toc
Joey Hess [Mon, 2 Sep 2024 20:31:31 +0000 (16:31 -0400)]
Merge branch 'master' of ssh://git-annex.branchable.com
Joey Hess [Mon, 2 Sep 2024 20:31:03 +0000 (16:31 -0400)]
fix number of headers
Joey Hess [Sat, 31 Aug 2024 23:50:36 +0000 (19:50 -0400)]
add news item for git-annex 10.
20240831
Joey Hess [Sat, 31 Aug 2024 23:50:26 +0000 (19:50 -0400)]
releasing package git-annex version 10.
20240831
Joey Hess [Fri, 30 Aug 2024 18:56:38 +0000 (14:56 -0400)]
use Utility.PID
fixes build on i386ancient
Joey Hess [Fri, 30 Aug 2024 18:49:18 +0000 (14:49 -0400)]
fix liveupdate locking
This fixes the build on windows.
Changed it to use lock pools, which will behave better if two threads
call getLiveRepoSizes at the same time.
Also this should make it work when annex.pidlock is set. In that case,
once the current process locks this file, or anything, any other process
will have to wait on the pid lock. So checkStaleSizeChanges will
correctly identify any other live changes in the database as stale,
since there can only be one git-annex process running.
Joey Hess [Fri, 30 Aug 2024 18:35:05 +0000 (14:35 -0400)]
fix paste oops
Wow, I pasted a big thing into entirely the wrong file, but it was in a
comment so it compiled anyway.
Joey Hess [Fri, 30 Aug 2024 16:06:45 +0000 (12:06 -0400)]
mention sizebalanced as well as balanced
Joey Hess [Fri, 30 Aug 2024 15:58:10 +0000 (11:58 -0400)]
avoid locking the journal in readonly repository
The test suite flagged that git-annex info in a readonly repository was
no longer working.
.git/annex/journal.lck: openFd: permission denied
This fixes it, however, in a case where .git/annex/reposize/ is
writable, but .git/annex/journal/ is not, there will still be a
permission denied error. The solution would just be to use consistent
permissions I suppose.
Joey Hess [Fri, 30 Aug 2024 15:14:45 +0000 (11:14 -0400)]
update
Joey Hess [Fri, 30 Aug 2024 15:08:32 +0000 (11:08 -0400)]
document using balanced preferred content in a cluster
Joey Hess [Fri, 30 Aug 2024 15:01:39 +0000 (11:01 -0400)]
Merge branch 'master' into balanced
Joey Hess [Fri, 30 Aug 2024 14:48:10 +0000 (10:48 -0400)]
lookupkey: Allow using --ref in a bare repository.
yarikoptic [Thu, 29 Aug 2024 18:35:47 +0000 (18:35 +0000)]
Added a comment
Joey Hess [Wed, 28 Aug 2024 18:13:12 +0000 (14:13 -0400)]
err on the side of larger repository size
When a live update is removing a key, it might fail. So only count those
once they have succeeded. When a live update is adding a key, count it
immediately to avoid over-filling a repo.
This also makes the 1 minute delay between stale live changes checks
more defensible, because a stale live change can only cause us to err
more on the side of caution.
Joey Hess [Wed, 28 Aug 2024 17:52:59 +0000 (13:52 -0400)]
remove stale live changes from reposize database
Reorganized the reposize database directory, and split up a column.
checkStaleSizeChanges needs to run before needLiveUpdate,
otherwise the process won't be holding a lock on its pid file, and
another process could go in and expire the live update it records. It
just so happens that they do get called in the correct order, since
checking balanced preferred content calls getLiveRepoSizes before
needLiveUpdate.
The 1 minute delay between checks is arbitrary, but will avoid excess
work. The downside of it is that, if a process is dropping a file and
gets interrupted, for 1 minute another process can expect a repository
will soon be smaller than it is. And so a process might send data to a
repository when a file is not really going to be dropped from it. But
note that can already happen if a drop takes some time in eg locking and
then fails. So it seems possible that live updates should only be
allowed to increase, rather than decrease the size of a repository.
Joey Hess [Wed, 28 Aug 2024 15:00:59 +0000 (11:00 -0400)]
combine 2 queries
Joey Hess [Wed, 28 Aug 2024 14:52:34 +0000 (10:52 -0400)]
avoid reposize database locking overhead when not needed
Only when the preferred content expression being matched uses balanced
preferred content is this overhead needed.
It might be possible to eliminate the locking entirely. Eg, check the
live changes before and after the action and re-run if they are not
stable. For now, this is good enough, it avoids existing preferred
content getting slow. If balanced preferred content turns out to be too
slow to check, that could be tried later.
matrss [Wed, 28 Aug 2024 14:11:36 +0000 (14:11 +0000)]
Added a comment
mih [Wed, 28 Aug 2024 13:58:30 +0000 (13:58 +0000)]
Added a comment: Needed to retrieve single file metadata from bare repo
matrss [Wed, 28 Aug 2024 08:47:33 +0000 (08:47 +0000)]
Added a comment
Joey Hess [Tue, 27 Aug 2024 19:47:57 +0000 (15:47 -0400)]
fix a deadlock when not using --auto
Live update never gets started, but then it still waited for it to
finish.
This only deadlocked with -J4 or so, not without -J. Unsure why.
Joey Hess [Tue, 27 Aug 2024 19:00:10 +0000 (15:00 -0400)]
avoid nub
There will not usually be many live changes, but usually does not mean
ever, and O(N^2) is best avoided.
Joey Hess [Tue, 27 Aug 2024 18:59:13 +0000 (14:59 -0400)]
thoughts
Joey Hess [Tue, 27 Aug 2024 17:07:06 +0000 (13:07 -0400)]
locking in checkLiveUpdate
This makes sure that two threads don't check balanced preferred content at the
same time, so each thread always sees a consistent picture of what is
happening.
This does add a fairly expensive file level lock to every check of
preferred content, in commands that use prepareLiveUpdate. It would
be good to only do that when live updates are actually needed, eg when
the preferred content expression uses balanced preferred content.
Joey Hess [Tue, 27 Aug 2024 15:04:27 +0000 (11:04 -0400)]
closing in on finishing live reposizes
Fixed successfullyFinishedLiveSizeChange to not update the rolling total
when a redundant change is in RecentChanges.
Made setRepoSizes clear RecentChanges that are no longer needed.
It might be possible to clear those earlier, this is only a convenient
point to do it.
The reason it's safe to clear RecentChanges here is that, in order for a
live update to call successfullyFinishedLiveSizeChange, a change must be
made to a location log. If a RecentChange gets cleared, and just after
that a new live update is started, making the same change, the location
log has already been changed (since the RecentChange exists), and
so when the live update succeeds, it won't call
successfullyFinishedLiveSizeChange. The reason it doesn't
clear RecentChanges when there is a reduntant live update is because
I didn't want to think through whether or not all races are avoided in
that case.
The rolling total in SizeChanges is never cleared. Instead,
calcJournalledRepoSizes gets the initial value of it, and then
getLiveRepoSizes subtracts that initial value from the current value.
Since the rolling total can only be updated by updateRepoSize,
which is called with the journal locked, locking the journal in
calcJournalledRepoSizes ensures that the database does not change while
reading the journal.
Joey Hess [Tue, 27 Aug 2024 14:17:43 +0000 (10:17 -0400)]
use live reposizes in balanced preferred content
Joey Hess [Tue, 27 Aug 2024 13:18:25 +0000 (09:18 -0400)]
fixed the build
Manually tested getLiveRepoSizes and it is working correctly.
Joey Hess [Tue, 27 Aug 2024 04:13:54 +0000 (00:13 -0400)]
fix a deadlock
When finishedLiveUpdate was run on a different key than expected, it
blocked forever waiting for an indication the database had been updated.
Since the journal is locked when finishedLiveUpdate runs, this could
also have caused other git-annex commands to hang.
Spencer [Mon, 26 Aug 2024 20:02:03 +0000 (20:02 +0000)]
Added contributions section to track my bugs and inquiries
Joey Hess [Mon, 26 Aug 2024 18:50:09 +0000 (14:50 -0400)]
started work on getLiveRepoSizes
Doesn't quite compile
Joey Hess [Mon, 26 Aug 2024 13:43:32 +0000 (09:43 -0400)]
partially fix concurrency issue in updating the rollingtotal
It's possible for two processes or threads to both be doing the same
operation at the same time. Eg, both dropping the same key. If one
finishes and updates the rollingtotal, then the other one needs to be
prevented from later updating the rollingtotal as well. And they could
finish at the same time, or with some time in between.
Addressed this by making updateRepoSize be called with the journal
locked, and only once it's been determined that there is an actual
location change to record in the log. updateRepoSize waits for the
database to be updated.
When there is a redundant operation, updateRepoSize won't be called,
and the redundant LiveUpdate will be removed from the database on
garbage collection.
But: There will be a window where the redundant LiveUpdate is still
visible in the db, and processes can see it, combine it with the
rollingtotal, and arrive at the wrong size. This is a small window, but
it still ought to be addressed. Unsure if it would always be safe to
remove the redundant LiveUpdate? Consider the case where two drops and a
get are all running concurrently somehow, and the order they finish is
[drop, get, drop]. The second drop seems redundant to the first, but
it would not be safe to remove it. While this seems unlikely, it's hard
to rule out that a get and drop at different stages can both be running
at the same time.
Joey Hess [Sun, 25 Aug 2024 14:48:42 +0000 (10:48 -0400)]
todo
Joey Hess [Sun, 25 Aug 2024 14:34:47 +0000 (10:34 -0400)]
rolling total of size changes in RepoSize database
When a live size change completes successfully, the same transaction
that removes it from the database updates the rolling total for its
repository.
The idea is that when RepoSizes is read, SizeChanges will be as
well, and cached locally. Any time a change is made, the local cache
will be updated. So by comparing the local cache with the current
SizeChanges, it can learn about size changes that were made by other
processes. Then read the LiveSizeChanges, and add that in to get a live
picture of the current sizes.
Also added a SizeChangeId. This allows 2 different threads, or
processes, to both record a live size change for the same repo and key,
and update their own information without stepping on one-another's toes.
Joey Hess [Sun, 25 Aug 2024 12:22:40 +0000 (08:22 -0400)]
use FileSize
It's just an alias, so this doesn't change the db schema, but it makes
explicit that it's not stored as an int64
Joey Hess [Sat, 24 Aug 2024 19:06:00 +0000 (15:06 -0400)]
update
Joey Hess [Sat, 24 Aug 2024 17:15:39 +0000 (13:15 -0400)]
update
Joey Hess [Sat, 24 Aug 2024 17:07:05 +0000 (13:07 -0400)]
improve live update starting
In an expression like "balanced=foo and exclude=bar", avoid it starting
a live update when the overall expression doesn't match.
Joey Hess [Sat, 24 Aug 2024 15:58:17 +0000 (11:58 -0400)]
todo
Joey Hess [Sat, 24 Aug 2024 15:49:58 +0000 (11:49 -0400)]
LiveUpdate db updates working
I've tested the behavior of the thread that waits for the LiveUpdate to
be finished, and it does get signaled and exit cleanly when the
LiveUpdate is GCed instead.
Made finishedLiveUpdate wait for the thread to finish updating the
database.
There is a case where GC doesn't happen in time and the database is left
with a live update recorded in it. This should not be a problem as such
stale data can also happen when interrupted and will need to be detected
when loading the database.
Balanced preferred content expressions now call startLiveUpdate.