28d87ca62d
fix(ft-fs): move default config to the backend impl
2023-04-07 17:25:22 +03:00
258fabbf8b
fix(fs-exp): reply with error when listing failed everywhere
2023-04-07 17:25:22 +03:00
11edfc1c6a
fix(ft): comment out some clauses as "unreachable"
2023-04-07 17:25:22 +03:00
bef5cc9c0f
fix(fs-fold): avoid folding through symlinked directories
...
Also a testsuite that verifies multilevel fold behaviour.
2023-04-07 17:25:22 +03:00
6ad7ce55d2
fix(ft-conf): separate local segments storage settings
...
To make things less ambiguous.
2023-04-07 17:25:22 +03:00
45e3b62dc4
refactor(ft): prefer plain `files` over `exports` in generic APIs
...
So there's no more `exports` magically becomings `files` in the
REST API which was slightly confusing.
2023-04-07 17:25:22 +03:00
c24c7eca34
fix(ft-test): unbreak testcase by inhibiting local fs storage GC
2023-04-07 17:25:22 +03:00
23cd78b8d6
refactor(ft-asm): turn state data record into a map
...
Which should be more future-proof.
2023-04-07 17:25:22 +03:00
b6b044f429
feat(ft): move out exporter concept into dedicated modules
2023-04-07 17:25:22 +03:00
54e54cc63d
feat(api): provide utility to qualify API URIs
2023-04-07 17:25:22 +03:00
64f15f1fdb
fix(ft-gc): ensure directories of complete transfers are GCed
2023-04-07 17:25:22 +03:00
2707b4500f
feat(ft-fs): ensure filsystem-safe handling of client input
...
Also restrict the filenames clients may specify in a filemeta to some
safe subset, as a future proofing measure.
2023-04-07 17:25:22 +03:00
9aec01e7a3
fix(ft-asm): use regular map for meta fragments
...
There's actually no need to use an ordered data structure at all.
2023-04-07 17:25:22 +03:00
4132f5a5fb
feat(ft): introduce exporter concept in local storage backend
...
The exporter is responsible for keeping fully transferred and
successfully assembled files. This was on the local storage itself
before. This abstraction is needed to give us an ability to
support S3 destinations more easily, just by swapping the storage
exporter.
Also implement local filesystem exporter and reimplement parts of
the `emqx_ft` API on top of it.
2023-04-07 17:25:22 +03:00
2880c8f4eb
fix(ft): unwrap error details when reader fails to start
2023-04-07 17:25:22 +03:00
45f00e14a9
feat(ft): ensure that clientid is always binary
...
For the sake of simplicity (e.g. transfer ids are now easier to
compare).
2023-04-07 17:25:22 +03:00
4f2600b9f1
feat(ft-gc): treat all transfer as incomplete
...
Since the concept of _complete transfers_ is being split out into
the _export_ concept, we lose knowledge of completeness in the GC.
Instead of asking exporters for transfer statuses we just treat all
transfer as incomplete when GCing.
2023-04-07 17:25:22 +03:00
0d39546080
feat(wdgraph): add `fold/3` which folds over graph edges
2023-04-07 17:25:22 +03:00
0c821cd3bd
test(ft): inject configs through hocon subsystem
...
So that relevant parts of config would be initialized with defaults.
2023-04-07 17:25:22 +03:00
e1dc48fa2b
feat(fs-gc): wire gc up with emqx config
2023-04-07 17:25:22 +03:00
50c6eef2bc
fix(fs-gc): do not hide `enoent`s
...
Also use `file:read_link_info/2`, it actually fetches any file info
while also not following symlinks automatically, which is better for
GC usecases.
2023-04-07 17:25:22 +03:00
bcd2099ce1
fix(fs-gc): make deletion empty transfer directories safer
2023-04-07 17:25:22 +03:00
344799f100
fix(fs-gc): add testcase on incomplete transfers GC
...
And fix logic that made GC delete incomplete segments prematurely.
2023-04-07 17:25:22 +03:00
15dc7c3e84
fix(test): fix node helper usage
2023-04-07 17:25:22 +03:00
715816e67b
feat(ft): add GC logic and process for the FS storage backend
2023-04-07 17:25:22 +03:00
e9f98adca2
test: make content_gen emit meta with each chunk
...
Which will also contain total content size, to increase this tool's
utility.
2023-04-07 17:25:22 +03:00
788e76ed2d
test: make assembly proptests waste less memory
2023-04-07 17:25:22 +03:00
8e6f960c09
refactor(ft): employ `emqx_wdgraph` for coverage computation
...
Also describe how coverage problem maps to shortest path problem.
2023-04-07 17:25:22 +03:00
b189ee463c
feat: add weighted directional graph ADT with shortest path
...
Basically, separate what abstraction was in `emqx_ft_assembly` into
dedicated module with a compact interface and a basic testsuite.
2023-04-07 17:25:22 +03:00
130601376a
perf(ft-asm): express coverage through shortest path on graph
...
Coverage becomes the shortest path problem on graph where nodes are
offsets and edges are nodes' segments. Implement a simple Dijkstra
algorithm to find one.
2023-04-07 17:25:22 +03:00
0c84fc28b0
perf(asm-ft): tradeoff optimality for computational complexity
...
Through squashing segments table into consecutive "runs".
2023-04-07 17:25:22 +03:00
97cfdf8eef
test(ft-asm): add property tests for file assembly
2023-04-07 17:25:22 +03:00
0af7e2a002
perf(ft-asm): optimize away remote segments identical to local
2023-04-07 17:25:22 +03:00
ccba65e1ea
chore(ft): add code owners for ft
2023-04-07 17:25:22 +03:00
15d967459c
feat(ft): add segment checksum validation
...
Also downgrade validation errors to mere info messages.
2023-04-07 17:25:22 +03:00
5998961f9f
fix(ft): log errors where they might get lost
2023-04-07 17:25:22 +03:00
5014cc15f4
test(ft): do not rely on fs backend implementation details
2023-04-07 17:25:22 +03:00
93865a79e9
fix(ft): disallow empty fileids
2023-04-07 17:25:22 +03:00
75070102ec
fix(ft): improve typespecs
2023-04-07 17:25:22 +03:00
2b925aa60b
fix(ft): drop unrelated TODO
2023-04-07 17:25:22 +03:00
7ff253e80d
test(ft): attempt to avoid `gen_rpc` config reset hacks
2023-04-07 17:25:22 +03:00
a3f47641f3
test(ft): describe the complex test case with comments
...
In order to make the test client model and behavior clearer.
2023-04-07 17:25:22 +03:00
10d4c4305a
feat(maybe): add basic tests for the new module
2023-04-07 17:25:22 +03:00
2cf2a2d952
fix(ft): make `fin` packet parser stricter + safer
2023-04-07 17:25:22 +03:00
58115715dd
fix(ft): require final size in `fin` packet
...
Otherwise there are situations when it's not entirely clear if a
transfer is really ready to be assembled. Since the `size` field
in a filemeta is not required (and rightly so), we need client to
tell us the final transfer size at the end of the process.
Also synthesize a testcase to show why it's needed.
Also worth noting that right now `fin` packets require final size,
even if a client already told us the size through filemeta. The
latter is regarded as serving informational purposes only (which
means that, for example, it might differ from the final size, or
some tranfer progress might show >100% somewhere because of that).
2023-04-07 17:25:22 +03:00
c073914f75
chore(ft): fix gen_rpc flakyness
2023-04-07 17:25:22 +03:00
228bf1a0ce
chore(ft): fix typing issues
2023-04-07 17:25:22 +03:00
2c9cd1397d
chore(ft): fix `?tp` calls
2023-04-07 17:25:22 +03:00
1c61165a91
fix(ft): fix emqx_cm takover
2023-04-07 17:25:22 +03:00
c6b3f2c2ca
feat(ft): add schema descriptions
2023-04-07 17:25:21 +03:00
Andrew Mayorov