If we fail to take the rename shortcut we may crash on a later loop,
because we do trickiness with the indexes but the original buckets[key]
in "range buckets[key]" isn't re-evaluated so i exceeds the max index.
This adds indirection of large version vectors in the same manner as we
already to block lists. The effect is the same: less duplicated data in
some situations.
To mitigate the impact for when this indirection
wouldn't be needed I've added an indirection cutoff for both blocks and
the new version vector stuff: we don't do the indirection at all for
small block lists or small version vectors, instead storing it directly
like we used to do. This is faster for small files and small setups.
This makes version vector values clock based instead of just incremented
from zero. The effect is that a vector that is created from scratch
(after database reset) will have a higher value for the local device
than what it could have been previously, causing a conflict. That is, if
we are A and we had
{A: 42, B: 12}
in the old scheme, a reset and rescan would give us
{A: 1}
which is a strict ancestor of the older file (this might be wrong). With
the new scheme we would instead have
{A: someClockTime, b: otherClockTime}
and the new version after reset would become
{A: someClockTime+delta}
which is in conflict with the previous entry (better).
In case the clocks are wrong (current time is less than the value in the
vector) we fall back to just simple increment like today.
This scheme is ineffective if we suffer a database reset while at the
same time setting the clock back far into the past. It's however no
worse than what we already do.
This loses the ability to emit the "added" event, as we can't look for
the magic 1 entry any more. That event was however already broken
(#5541).
Another place where we infer meaning from the vector itself is in
receive only folders, but there the only criteria is that the vector is
one item long and includes just ourselves, which remains the case with
this change.
* wip
- In the few places where we wrap errors, use the new Go 1.13 "%w"
construction instead of %s or %v.
- Where we create errors with constant strings, consistently use
errors.New and not fmt.Errorf.
- Remove capitalization from errors in the few places where we had that.
One of the causes of "panic: database is closed" is that we try to send
summaries after it's been closed. Calculating summaries can take a long
time and if we have a lot of folders it's not unreasonable to think
that we might be stopped in this loop, so prepare to bail here.
* push
During some other work I discovered these tests weren't great, so I've
rewritten them to be a little better. The real changes here are:
- Don't play games with not starting the folder and such, and don't
construct a fake folder instance -- just use the one the model has. The
folder starts and scans but the folder contents are empty at this point
so that's fine.
- Use a fakefs instead of a temp dir.
- To support the above, implement a fakefs option `?content=true` to
make the fakefs actually retain written content. Use sparingly,
obviously, but it means the fakefs can usually be used instead of an
on disk real directory.
This adds a new config with the simple and concise name
maxConcurrentIncomingRequestKiB. This limits how many bytes we have "in
the air" in the form of response data being read and processed.
After some testing I think that not having this limiter is seldom a
great idea and thus I propose a default value of 256 MiB for this new
setting.
I also refactored the folder IO limiter to be a model/folder attribute
instead of a package global.
Adds a new folder state "Waiting to Sync" in the same vein as the
existing "Waiting to Scan". This vastly improves performances in the
rare cases when there are lots and lots of folders operating.
* lib/db: Deduplicate block lists in database (fixes#5898)
This moves the block list in the database out from being just a field on
the FileInfo to being an object of its own. When putting a FileInfo we
marshal the block list separately and store it keyed by the sha256 of
the marshalled block list. When getting, if we are not doing a
"truncated" get, we do an extra read and unmarshal for the block list.
Old block lists are cleared out by a periodic GC sweep. The alternative
would be to use refcounting, but:
- There is a larger risk of getting that wrong and either dropping a
block list in error or keeping them around forever.
- It's tricky with our current database, as we don't have dirty reads.
This means that if we update two FileInfos with identical block lists in
the same transaction we can't just do read/modify/write for the ref
counters as we wouldn't see our own first update. See above about
tracking this and risks about getting it wrong.
GC uses a bloom filter for keys to avoid heavy RAM usage. GC can't run
concurrently with FileInfo updates so there is a new lock around those
operation at the lowlevel.
The end result is a much more compact database, especially for setups
with many peers where files get duplicated many times.
This is per-key-class stats for a large database I'm currently working
with, under the current schema:
```
0x00: 9138161 items, 870876 KB keys + 7397482 KB data, 95 B + 809 B avg, 1637651 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x08: 1349 items, 12 KB keys + 10 KB data, 9 B + 8 B avg, 17 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 7 B avg, 30 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
Total 10426475 items, 968490 KB keys + 9202925 KB data.
```
Note 7.4 GB of data in class 00, total size 9.2 GB. After running the
migration we get this instead:
```
0x00: 9138161 items, 870876 KB keys + 2611392 KB data, 95 B + 285 B avg, 4788 B max
0x01: 185656 items, 10388 KB keys + 1790909 KB data, 55 B + 9646 B avg, 924525 B max
0x02: 916890 items, 84795 KB keys + 3667 KB data, 92 B + 4 B avg, 192 B max
0x03: 384 items, 27 KB keys + 5 KB data, 72 B + 15 B avg, 87 B max
0x04: 1109 items, 17 KB keys + 17 KB data, 15 B + 15 B avg, 69 B max
0x06: 383 items, 3 KB keys + 0 KB data, 9 B + 2 B avg, 18 B max
0x07: 510 items, 4 KB keys + 12 KB data, 9 B + 24 B avg, 41 B max
0x09: 194 items, 0 KB keys + 123 KB data, 5 B + 634 B avg, 11484 B max
0x0a: 3 items, 0 KB keys + 0 KB data, 14 B + 17 B avg, 51 B max
0x0b: 181836 items, 2363 KB keys + 10694 KB data, 13 B + 58 B avg, 173 B max
0x0d: 44282 items, 1461 KB keys + 61081 KB data, 33 B + 1379 B avg, 1637399 B max
Total 10469408 items, 969939 KB keys + 4477905 KB data.
```
Class 00 is now down to 2.6 GB, with just 61 MB added in class 0d.
There will be some additional reads in some cases which theoretically
hurts performance, but this will be more than compensated for by smaller
writes and better compaction.
On my own home setup which just has three devices and a handful of
folders the difference is smaller in absolute numbers of course, but
still less than half the old size:
```
0x00: 297122 items, 20894 KB keys + 306860 KB data, 70 B + 1032 B avg, 103237 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
Total 1947412 items, 151268 KB keys + 337485 KB data.
```
to:
```
0x00: 297122 items, 20894 KB keys + 37038 KB data, 70 B + 124 B avg, 520 B max
0x01: 115299 items, 7738 KB keys + 17542 KB data, 67 B + 152 B avg, 419 B max
0x02: 1430537 items, 121223 KB keys + 5722 KB data, 84 B + 4 B avg, 253 B max
...
0x0d: 18041 items, 595 KB keys + 71964 KB data, 33 B + 3988 B avg, 101109 B max
Total 1965447 items, 151863 KB keys + 139628 KB data.
```
* wip
* wip
* wip
* wip
Makes the logic a bit clearer and safer. This also sneakily redefines
the 0 interval to also mean disabled, whereas it previously meant ...
sometimes default to 1s, sometimes just spin.
This adds error returns to model methods called by the protocol layer.
Returning an error will cause the connection to be torn down as the
message couldn't be handled. Using this to signal that a folder isn't
currently available will then cause a reconnection a few moments later,
when it'll hopefully work better.
Tested manually by running with STRECHECKDBEVERY=0 on a nontrivially
sized setup. This panics reliably before this patch, but just causes a
disconnect/reconnect now.
As foretold by the prophecy, "once the database refactor is merged, then
shall appear a request to propagate errors from the store known
throughout the land as the NamedspacedKV, and it shall be good".
This PR does two things, because one lead to the other:
- Move the leveldb specific stuff into a small "backend" package that
defines a backend interface and the leveldb implementation. This allows,
potentially, in the future, switching the db implementation so another
KV store should we wish to do so.
- Add proper error handling all along the way. The db and backend
packages are now errcheck clean. However, I drew the line at modifying
the FileSet API in order to keep this manageable and not continue
refactoring all of the rest of Syncthing. As such, the FileSet methods
still panic on database errors, except for the "database is closed"
error which is instead handled by silently returning as quickly as
possible, with the assumption that we're anyway "on the way out".
* lib/versioner: Reduce surface area
This is a refactor while I was anyway rooting around in the versioner.
Instead of exporting every possible implementation and the factory and
letting the caller do whatever, this now encapsulates all that and
exposes a New() that takes a config.VersioningConfiguration.
Given that and that we don't know (from the outside) how a versioner
works or what state it keeps, we now just construct it once per folder
and keep it around. Previously it was recreated for each restore
request.
* unparam
* wip
Assume a folder error was set due to bad ignores on the latest scan.
Previously, doing a manual rescan would result in:
1. Clearing the folder error, which schedules (immediately) an fs
watcher restart
2. Attempting to load the ignores, which fails, so we set a folder
error and bail.
3. Now the fs watcher restarts, as scheduled, so we trigger a scan.
Goto 1.
This change fixes this by not clearing the error until the error is
actually cleared, that is, if both the health check and ignore loading
succeeds.
* lib/model: Don't panic on failed chmod-back on directory (fixes#5836)
This makes the "in writable dir"-wrapper log chmod-back errors instead
of panicking. To do that we need a logger so the function moved into the
model package which is also the only place it's used. The tests came
along.
(The test also exercised osutil.RenameOrCopy like some sort of
piggybacking. I removed that.)
* lib/fs, lib/model: Add error channel to Watch to avoid panics (fixes#5697)
* forgot unsupported watch
* and more non(-standard)-unixy fixes
* and windows test
* review
The check in ClusterConfig() when iterating through announced devices
in a folder explicitly skips entries without a non-zero IndexID.
Therefore, the check for IndexID == 0 just below will never be true
and the intended cleanup of local index data will not happen.
Plainly remove that check to make the intended case distinction work.
* lib/protocol: Wait for reader/writer loops on close (fixes#4170)
* waitgroup
* lib/model: Don't hold lock while closing connection
* fix comments
* review (lock once, func argument) and naming
