This adds pre-validation directly at the unmarshal step, of annotations
added to the protobuf definitions. This seeks to weed out unreasonable
situations early, by ensuring offsets are positive, string lengths are
limited, etc. Some of the limits are a bit subjective, but I think I've
put them above what is expected and reasonable...
---------
Signed-off-by: Jakob Borg <[email protected]>
The API can use arbitrary paging parameters, causing arbitrary numbers
of prepared statements, which becomes a memory leak.
Signed-off-by: Jakob Borg <[email protected]>
This could happen when a folder has deleted files as the latest entries,
and nothing else happened for 180 days (the default tombstone cleanup
time).
Signed-off-by: Jakob Borg <[email protected]>
There is a danger of deadlock wherever we have database operations
inside a database iterator. The iterator itself pins a connection, so
the nested operations need another connection; if we've reached
maxOpenConns then it blocks until a connection is released. If all
connections are consumed by such iterators then none can make progress.
Luckily, we have limited concurrency for most such iterator loops. They
are part of scanning, pulling, reverting, etc where there is only ever
one such routine per folder. The exception is block reuse in the copier
routine, which is limited by the `Copiers` setting per folder. This
could in practice deadlock since you could set copiers to eight and end
up with six `AllLocalBlocksWithHash` iterators when `maxDBConns=6`, all
of which need to make additional database calls inside the loop.
This PR fixes the problem twice;
- The problematic loop does not need to be reentrant. The set of blocks
that may be returned by the iterator is finite so we can easily just
collect them to a slice before we start processing them. This avoids the
problem entirely.
- We do not need to limit database connections as strictly as we
currently do. Increase the maximum allowed, while reducing the number of
held-open idle connections slightly. This is not an exact science, but
ideally we want "most" operations to be able to use the pinned
connections to avoid cache churn. Most operations are short lived
queries, or single-goroutine iterators with short lived queries inside,
so four connections seems like it should usually be enough. 🤷
- Additionally, set a cap on Copiers. Currently you could set it to an
arbitrarily large number, which is not advantageous. Limit it to
2*NumCPU which scales somewhat with system performance.
Signed-off-by: Jakob Borg <[email protected]>
The counts needs to be modified manually since we're not running
triggers during migrations. The schema version needed to be bumped.
Signed-off-by: Jakob Borg <[email protected]>
Resp. directories for the database, and both dirs and symlinks for
`FileInfo.FileSize`. Instead handle it in the UI progress percentage. We
even already have special cases there for deletions, might as well
handle directories just like any other zero-sized needed item there.
This went through the very thorough testing of running it on my laptop,
the migration was applied and it seemed to be working fine after.
---------
Signed-off-by: Simon Frei <[email protected]>
### Purpose
`path` is for slash-separated paths (URLs, BEP protocol); local file
system paths should use `path/filepath`. Fixed in
`cmd/stdiscosrv/database.go` (3 sites) and
`internal/db/sqlite/db_test.go` (1 site).
### Testing
`go build ./cmd/stdiscosrv/...` and `go vet` pass.
Signed-off-by: Yasuhiro Matsumoto <[email protected]>
This adds a new folder-level configuration `FullBlockIndex`. It controls
whether we maintain the block index for a given folder -- currently
that's always true, now it becomes possible to turn off. The block index
is used for lookup of blocks across files and folders. Effectively, when
syncing a change, for each block, we check:
1. Is the block already present in the old version of the file? If so,
we can reuse (copy) it without network transfer. **This check is always
possible.**
2. Is the block already present in any other file in this folder or
other folders? If so we can copy it. **This check is only possible with
the full block index.**
3. We must transfer the block over the network.
Maintaining the full block index is costly in time, I/O and database
size. With this PR, maintaining the full block index becomes the default
for send-receive and receive-only folders only, with it disabled for
send-only and receive-encrypted folders. The block index is never useful
for encrypted folders, as blocks are encrypted separate for each file.
It is also not useful for send-only folders by themselves, though the
data in the send-only folder could be reused by other receive-type
folders if it were enabled.
For very large folders it may make sense to disable the full block index
regardless of folder type and just accept the resulting decrease in data
reuse.
Disabling or enabling the option in the GUI causes the index to be
destroyed or rebuilt accordingly.
https://github.com/syncthing/docs/pull/1005
---------
Signed-off-by: Jakob Borg <[email protected]>
chore(sqlite): reduce max open connections, keep them open permanently (fixes#10592)
Reduces connection churn, possibly tickling concurrency bug on Windows.
Signed-off-by: Jakob Borg <[email protected]>
Also adds a method to query the last database maintenance time.
Signed-off-by: Tommy van der Vorst <[email protected]>
Co-authored-by: Jakob Borg <[email protected]>
This change allows the periodic database maintenance to be disabled, while providing a way to programmatically start maintenance at a convenient moment.
Signed-off-by: Tommy van der Vorst <[email protected]>
Due to a thinko, this optimisation was wildly incorrect and would read
to lack of block reuse when syncing files.
(We do not insert a blocklist per device, but only a single one. We
can't use the fact of whether the insert happened as a criteria for
inserting blocks.)
Signed-off-by: Jakob Borg <[email protected]>
This adds a new field to the file information we keep, the "previous
blocks hash". This is the hash of the file contents as it was in its
previous incarnation. That is, every scan that updates the blocks hash
will move the current hash to the "previous" field.
This enables an addition to the conflict detection algorithm: if the
file to be synced is in conflict with the current file on disk
(version-counter wise), but it indicates that it was based on the
precise contents we have (new.prevBlocksHash == current.blocksHash),
then it's not really a conflict.
Signed-off-by: Jakob Borg <[email protected]>
This changes the files table to use normalisation for the names and
versions. The idea is that these are often common between all remote
devices, and repeating an integer is more efficient than repeating a
long string. A new benchmark bears this out; for a database with 100k
files shared between 31 devices, with some worst case assumption on
version vector size, the database is reduced in size by 50% and the test
finishes quicker:
Current:
db_bench_test.go:322: Total size: 6263.70 MiB
--- PASS: TestBenchmarkSizeManyFilesRemotes (1084.89s)
New:
db_bench_test.go:326: Total size: 3049.95 MiB
--- PASS: TestBenchmarkSizeManyFilesRemotes (776.97s)
The other benchmarks end up about the same within the margin of
variability, with one possible exception being that RemoteNeed seems to
be a little slower on average:
old files/s new files/s
Update/n=RemoteNeed/size=1000-8 5.051k 4.654k
Update/n=RemoteNeed/size=2000-8 5.201k 4.384k
Update/n=RemoteNeed/size=4000-8 4.943k 4.242k
Update/n=RemoteNeed/size=8000-8 5.099k 3.527k
Update/n=RemoteNeed/size=16000-8 3.686k 3.847k
Update/n=RemoteNeed/size=30000-8 4.456k 3.482k
I'm not sure why, possibly that query can be optimised anyhow.
Signed-off-by: Jakob Borg <[email protected]>
Store the sequence number of the last GC sweep in a KV. Next time, if it
matches we can just skip GC because nothing has been added or removed.
Signed-off-by: Jakob Borg <[email protected]>
Periodic garbage collection can take a long time on large folders. The worst
step is the one for blocks, which are typically orders of magnitude more
numerous than files or block lists.
This improves the situation in by running blocks GC in a number of smaller
range chunks, in random order, and stopping after a time limit. At most ten
minutes per run will be spent garbage collecting blocklists and blocks.
With this, we're not guaranteed to complete a full GC on every run, but
we'll make some progress and get there eventually.
Signed-off-by: Jakob Borg <[email protected]>
No migration on this as it has no practical impact, just a slight
cleanup for new installations.
Also a refactor of how we declare single column primary keys, for
consistency.
Signed-off-by: Jakob Borg <[email protected]>
This adds several options for configuring the log format of timestamps
and severity levels, making it more suitable for integration with log
systems like systemd.
--log-format-timestamp="2006-01-02 15:04:05"
Format for timestamp, set to empty to disable timestamps ($STLOGFORMATTIMESTAMP)
--[no-]log-format-level-string
Whether to include level string in log line ($STLOGFORMATLEVELSTRING)
--[no-]log-format-level-syslog
Whether to include level as syslog prefix in log line ($STLOGFORMATLEVELSYSLOG)
So, to get a timestamp suitable for systemd (syslog prefix, no level
string, no timestamp) we can pass `--log-format-timestamp=""
--no-log-format-level-string --log-format-level-syslog` or,
equivalently, set `STLOGFORMATTIMESTAMP="" STLOGFORMATLEVELSTRING=false
STLOGFORMATLEVELSYSLOG=true`.
Signed-off-by: Jakob Borg <[email protected]>
No practical effect, just a tiny bit of fun to stamp the database files
with an application ID that identifies them.
Signed-off-by: Jakob Borg <[email protected]>
While we're figuring out optimal defaults, reduce the page cache size to
the compiled-in default. In my computer this makes no difference in
benchmarks. In forum threads, it solved the problem of massive memory
usage during initial scan.
Signed-off-by: Jakob Borg <[email protected]>
This reduces database migration memory usage in my test scenario from
3.8 GB to 440 MB. In principle I don't think we're causing many temp
tables to be generated anyway in normal usage, but if we do and someone
can benchmark a performance difference, we can add a tunable. I ran the
database benchmark before and after and didn't see a difference above
the noise level.
Signed-off-by: Jakob Borg <[email protected]>
Just to be entirely sure that if the migration succeeds the schema
version is always also updated. Currently if a migration succeeds but a
later migration doesn't, the changes of the migration apply but the
version stays - if the migration is breaking/non-idempotent, it will
fail when it tries to rerun it next time (otherwise it's just a
pointless re-execution).
Unfortunately with the current `db.runScripts` it wasn't that easy to
do, so I had to do quite a bit of refactoring. I am also ensuring the
right order of transactions now, though I assume that was already the
case lexicographically - can't hurt to be safe.
This just removes an unnecessary foreign key constraint, where we
already do the garbage collection manually in the database service.
However, as part of getting here I tried a couple of other variants
along the way:
- Changing the order of the primary key from `(hash, blocklist_hash,
idx)` to `(blocklist_hash, idx, hash)` so that inserts would be
naturally ordered. However this requires a new index `on blocks (hash)`
so that we can still look up blocks by hash, and turns out to be
strictly worse than what we already have.
- Removing the primary key entirely and the `WITHOUT ROWID` to make it a
rowid table without any required order, and an index as above. This is
faster when the table is small, but becomes slower when it's large (due
to dual indexes I guess).
These are the benchmark results from current `main`, the second
alternative below ("Index(hash)") and this proposal that retains the
combined primary key ("combined"). Overall it ends up being about 65%
faster.
<img width="764" height="452" alt="Screenshot 2025-08-29 at 14 36 28"
src="https://github.com/user-attachments/assets/bff3f9d1-916a-485f-91b7-b54b477f5aac"
/>
Ref #10264
This adds a cleanup stage to remove database files for folders that no
longer exist on startup. Folder database files were already removed when
dropping a folder, assuming that the folder database had been opened at
that point. This won't be the case though when a folder is removed from
the config when Syncthing isn't running, or when a folder is dropped and
re-migrated in a restarted migration.