Commit Graph

27 Commits

Author SHA1 Message Date
Jakob Borg
deb2f42a35 wip 2023-06-29 15:51:14 +02:00
Jakob Borg
2a77eb54ee scanner 2023-06-29 15:35:25 +02:00
Jakob Borg
ed588ce335
lib/scanner: More sensible debug output (#8596)
Previous debug input didn't really give enough info to show what was
happening, while it also printed full block lists which are enormously
verbose. Now it consistently prints 1. what it sees on disk, 2. what it
got from CurrentFile (without blocks), 3. the action taken on that file.
2022-10-13 19:32:58 +02:00
Simon Frei
78c2844e3f
lib/scanner: Pass on errors while hashing (#7380) 2021-02-19 08:51:39 +01:00
greatroar
42917d707d
lib/scanner: Remove unused field, move WaitGroup.Add out of loop (#7323) 2021-02-03 14:25:24 +01:00
Simon Frei
8452fd2ab4
lib/model, lib/scanner: Prevent races aborting scans (fixes #6994) (#6997) 2020-09-25 11:27:44 +02:00
Jakob Borg
8fc2dfad0c
lib/db: Deduplicate block lists in database (fixes #5898) (#6283)
* 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
2020-01-24 08:35:44 +01:00
Simon Frei
d510e3cca3 all: Display errors while scanning in web UI (fixes #4480) (#5215) 2018-11-07 11:04:41 +01:00
Audrius Butkevicius
ef0dcea6a4 lib/model: Verify request content against weak (and possibly strong) hash (#4767) 2018-05-05 10:24:44 +02:00
Jakob Borg
19c7cd99f5 all: Implement variable sized blocks (fixes #4807) 2018-04-16 19:08:50 +01:00
Simon Frei
68c1b2dd47 all: Revert simultaneously walk fs and db on scan (fixes #4756) (#4757)
This reverts commit 6d3f9d5154.
2018-02-14 08:59:46 +01:00
Simon Frei
6d3f9d5154
all: Simultaneously walk fs and db on scan (fixes #2571, fixes #4573) (#4584)
When scanner.Walk detects a change, it now returns the new file info as well as the old file info. It also finds deleted and ignored files while scanning.
Also directory deletions are now always committed to db after their children to prevent temporary failure on remote due to non-empty directory.
2018-02-10 16:56:53 +01:00
Audrius Butkevicius
3d8b4a42b7 all: Convert folders to use filesystem abstraction
GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/4228
2017-08-19 14:36:56 +00:00
Jakob Borg
d6fbfc3545 lib/fs, lib/model, lib/scanner: Make scans cancellable (fixes #3965)
The folder already knew how to stop properly, but the fs.Walk() didn't
and can potentially take a very long time. This adds context support to
Walk and the underlying scanning stuff, and passes in an appropriate
context from above. The stop channel in model.folder is replaced with a
context for this purpose.

To test I added an infiniteFS that represents a large amount of data
(not actually infinite, but close) and verify that walking it is
properly stopped. For that to be implemented smoothly I moved out the
Walk function to it's own type, as typically the implementer of a new
filesystem type might not need or want to reimplement Walk.

It's somewhat tricky to test that this actually works properly on the
actual sendReceiveFolder and so on, as those are started from inside the
model and the filesystem isn't easily pluggable etc. Instead I've tested
that part manually by adding a huge folder and verifying that pause,
resume and reconfig do the right things by looking at debug output.

GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/4117
2017-04-26 00:15:23 +00:00
Jakob Borg
4253f22680 lib/scanner: Use fs.Filesystem for all operations
One more step on the path of the great refactoring. Touches rwfolder a
little bit since it uses the Lstat from fs as well, but mostly this is
just on the scanner as rwfolder is scheduled for a later refactor.

There are a couple of usages of fs.DefaultFilesystem that will in the
end become a filesystem injected from the top, but that comes later.

GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/4070
LGTM: AudriusButkevicius, imsodin
2017-04-01 09:04:11 +00:00
Jakob Borg
f7fc0c1d3e all: Update license url to https (ref #3976) 2017-02-09 08:04:16 +01:00
Audrius Butkevicius
dd78177ae0 scanner: Allow disabling weak hash in scanning (fixes #3891)
GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3905
2017-01-23 13:50:32 +00:00
Jakob Borg
7aaa1dd8a3 lib/scanner: Recheck file size and modification time after hashing (ref #3440)
To catch the case where the file changed. Also make sure we never let a
size-vs-blocklist mismatch slip through.

GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3443
2016-07-26 08:51:39 +00:00
Jakob Borg
fa0101bd60 lib/protocol, lib/discover, lib/db: Use protocol buffer serialization (fixes #3080)
This changes the BEP protocol to use protocol buffer serialization
instead of XDR, and therefore also the database format. The local
discovery protocol is also updated to be protocol buffer format.

GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3276
LGTM: AudriusButkevicius
2016-07-04 10:40:29 +00:00
klemens
bd41e21c26 all: Correct spelling in comments
Skip-check: authors pr-build-mac

GitHub-Pull-Request: https://github.com/syncthing/syncthing/pull/3056
2016-05-08 10:54:22 +00:00
Jakob Borg
a8a2192cf9 Show scan rate in web GUI 2015-11-17 21:23:17 +01:00
Jakob Borg
9fbdb6b305 Cancel a running scan 2015-11-13 15:30:21 +01:00
Jakob Borg
76af9ba53d Implement facility based logger, debugging via REST API
This implements a new debug/trace infrastructure based on a slightly
hacked up logger. Instead of the traditional "if debug { ... }" I've
rewritten the logger to have no-op Debugln and Debugf, unless debugging
has been enabled for a given "facility". The "facility" is just a
string, typically a package name.

This will be slightly slower than before; but not that much as it's
mostly a function call that returns immediately. For the cases where it
matters (the Debugln takes a hex.Dump() of something for example, and
it's not in a very occasional "if err != nil" branch) there is an
l.ShouldDebug(facility) that is fast enough to be used like the old "if
debug".

The point of all this is that we can now toggle debugging for the
various packages on and off at runtime. There's a new method
/rest/system/debug that can be POSTed a set of facilities to enable and
disable debug for, or GET from to get a list of facilities with
descriptions and their current debug status.

Similarly a /rest/system/log?since=... can grab the latest log entries,
up to 250 of them (hardcoded constant in main.go) plus the initial few.

Not implemented in this commit (but planned) is a simple debug GUI
available on /debug that shows the current log in an easily pasteable
format and has checkboxes to enable the various debug facilities.

The debug instructions to a user then becomes "visit this URL, check
these boxes, reproduce your problem, copy and paste the log". The actual
log viewer on the hypothetical /debug URL can poll regularly for new log
entries and this bypass the 250 line limit.

The existing STTRACE=foo variable is still obeyed and just sets the
start state of the system.
2015-10-03 18:09:53 +02:00
Jakob Borg
4581c57478 Fix import paths 2015-09-22 19:38:46 +02:00
Jakob Borg
bc016e360e Refactor: ints used in arithmetic should be signed 2015-08-27 21:37:12 +02:00
AudriusButkevicius
94c52e3a77 Add scan percentages (fixes #1030) 2015-08-27 19:20:43 +01:00
Jakob Borg
7705a6c1f1 mv internal lib 2015-08-09 09:35:26 +02:00