Commit graph

3534 commits

Author SHA1 Message Date
a49f0717b3 [] metabase: Cache frequently accessed singleton buckets
There are some buckets we access almost always, to check whether an
object is alive. In search we also iterate over lots of objects, and
`tx.Bucket()` shows itself a lot in pprof.
```
goos: linux
goarch: amd64
pkg: git.frostfs.info/TrueCloudLab/frostfs-node/pkg/local_object_storage/metabase
cpu: 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
                          │      1      │                  2                   │
                          │   sec/op    │    sec/op     vs base                │
Select/string_equal-8       4.753m ± 6%   3.969m ± 14%  -16.50% (p=0.000 n=10)
Select/string_not_equal-8   4.247m ± 9%   3.486m ± 11%  -17.93% (p=0.000 n=10)
Select/common_prefix-8      4.163m ± 5%   3.323m ±  5%  -20.18% (p=0.000 n=10)
Select/unknown-8            3.557m ± 3%   3.064m ±  8%  -13.85% (p=0.001 n=10)
geomean                     4.158m        3.445m        -17.15%

                          │      1       │                  2                   │
                          │     B/op     │     B/op      vs base                │
Select/string_equal-8       2.250Mi ± 0%   1.907Mi ± 0%  -15.24% (p=0.000 n=10)
Select/string_not_equal-8   2.250Mi ± 0%   1.907Mi ± 0%  -15.24% (p=0.000 n=10)
Select/common_prefix-8      2.250Mi ± 0%   1.907Mi ± 0%  -15.24% (p=0.000 n=10)
Select/unknown-8            2.243Mi ± 0%   1.900Mi ± 0%  -15.29% (p=0.000 n=10)
geomean                     2.248Mi        1.905Mi       -15.26%

                          │      1      │                  2                  │
                          │  allocs/op  │  allocs/op   vs base                │
Select/string_equal-8       56.02k ± 0%   47.03k ± 0%  -16.05% (p=0.000 n=10)
Select/string_not_equal-8   56.02k ± 0%   47.03k ± 0%  -16.05% (p=0.000 n=10)
Select/common_prefix-8      56.02k ± 0%   47.03k ± 0%  -16.05% (p=0.000 n=10)
Select/unknown-8            55.03k ± 0%   46.04k ± 0%  -16.34% (p=0.000 n=10)
geomean                     55.78k        46.78k       -16.12%
```

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-20 10:17:42 +00:00
a7ac30da9c [] tree: Refactor getSortedSubTree
* Reuse `item` as result for `forest.TreeSortedByFilename`
  invocation.

Signed-off-by: Airat Arifullin <a.arifullin@yadro.com>
2025-03-20 10:12:49 +00:00
39f549a7ab [] tree: Intoduce a helper LastChild
Signed-off-by: Airat Arifullin <a.arifullin@yadro.com>
2025-03-20 10:12:49 +00:00
760b6a44ea [] tree: Fix sorted getSubtree for multiversion filenames
Signed-off-by: Airat Arifullin <a.arifullin@yadro.com>
2025-03-20 10:12:49 +00:00
a11b2d27e4 [] tree: Introduce Cursor type
* Use `Cursor` as parameter for `TreeSortedByFilename`

Signed-off-by: Airat Arifullin <a.arifullin@yadro.com>
2025-03-20 10:12:49 +00:00
a405fb1f39 [] metabase: Check object status once in Select()
objectStatus() is called twice for the same object:
First, in selectObject() to filter removed objects.
Then, again, in getObjectForSlowFilters() via db.get().
The second call will return the same result, so remove useless branch.

```
goos: linux
goarch: amd64
pkg: git.frostfs.info/TrueCloudLab/frostfs-node/pkg/local_object_storage/metabase
cpu: 11th Gen Intel(R) Core(TM) i5-1135G7 @ 2.40GHz
                          │     old     │                status                │
                          │   sec/op    │    sec/op     vs base                │
Select/string_equal-8       5.022m ± 7%   3.968m ±  8%  -20.98% (p=0.000 n=10)
Select/string_not_equal-8   4.953m ± 9%   3.990m ± 10%  -19.44% (p=0.000 n=10)
Select/common_prefix-8      4.962m ± 8%   3.971m ±  9%  -19.98% (p=0.000 n=10)
Select/unknown-8            5.246m ± 9%   3.548m ±  5%  -32.37% (p=0.000 n=10)
geomean                     5.045m        3.865m        -23.39%

                          │     old      │                status                │
                          │     B/op     │     B/op      vs base                │
Select/string_equal-8       2.685Mi ± 0%   2.250Mi ± 0%  -16.20% (p=0.000 n=10)
Select/string_not_equal-8   2.685Mi ± 0%   2.250Mi ± 0%  -16.20% (p=0.000 n=10)
Select/common_prefix-8      2.685Mi ± 0%   2.250Mi ± 0%  -16.20% (p=0.000 n=10)
Select/unknown-8            2.677Mi ± 0%   2.243Mi ± 0%  -16.24% (p=0.000 n=10)
geomean                     2.683Mi        2.248Mi       -16.21%

                          │     old     │               status                │
                          │  allocs/op  │  allocs/op   vs base                │
Select/string_equal-8       69.03k ± 0%   56.02k ± 0%  -18.84% (p=0.000 n=10)
Select/string_not_equal-8   69.03k ± 0%   56.02k ± 0%  -18.84% (p=0.000 n=10)
Select/common_prefix-8      69.03k ± 0%   56.02k ± 0%  -18.84% (p=0.000 n=10)
Select/unknown-8            68.03k ± 0%   55.03k ± 0%  -19.11% (p=0.000 n=10)
geomean                     68.78k        55.77k       -18.90%
```

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-18 11:48:51 +00:00
a7319bc979 [] metabase/test: Report allocs in benchmarkSelect()
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-18 11:48:51 +00:00
07a660fbc4
[] writecache: Add QoS limiter usage
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-14 16:23:33 +03:00
7893d763d1
[] logger: Add sampling for journald logger
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-13 14:52:03 +03:00
ff4e9b6ae1
[] logger: Drop unused fields
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-13 14:52:02 +03:00
997759994a
[] golangci: Enable gci linter
Signed-off-by: Alexander Chuprov <a.chuprov@yadro.com>
2025-03-13 12:04:01 +03:00
ecb6b0793c [] Use slices.ContainsFunc() where possible
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-13 08:12:20 +00:00
460e5cbccf [] Use slices.Delete() where possible
gopatch is missing for this one, because
https://github.com/uber-go/gopatch/issues/179

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-13 08:12:20 +00:00
155d3ddb6e [] Use min builtin where possible
Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-13 08:12:20 +00:00
40536d8a06 [] Use fmt.Appendf where warranted
Fix gopls warnings:
```
cmd/frostfs-adm/internal/modules/morph/config/config.go:68:20-64: Replace []byte(fmt.Sprintf...) with fmt.Appendf
````

gopatch:
```
@@
var f expression
@@
-[]byte(fmt.Sprintf(f, ...))
+fmt.Appendf(nil, f, ...)
```

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-13 08:12:20 +00:00
bcc84c85a0 [] Replace interface{} with any
gopatch:
```
@@
@@
-interface{}
+any
```

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-13 08:12:20 +00:00
2005fdda09
[] shard: Drop shard pool
After adding an ops limiter, shard's `put` pool is redundant.

Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-11 13:59:51 +03:00
597bce7a87 [] treeSvc: Add operations by IO tag metric
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-11 10:57:47 +00:00
4ed2bbdb0f [] objectSvc: Add operations by IO tag metric
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-11 10:57:47 +00:00
3727d60331 [] qos: Add metrics
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-11 10:57:47 +00:00
8643e0abc5
[] writecache: Use object size to check free space
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-03-10 17:52:57 +03:00
bd61f7bf0a [] audit: Fix duplicated log in Patch method
When we do `object patch` with audit enabled we get several
duplicated entries in logs.

`object patch` request is logged in 2 places:
1. `(*auditPatchStream) CloseAndRecv()` - when the client closes
   the request stream or when stream gets aborted.
2. `(*auditPatchStream) Send()` - when stream was NOT aborted.

`Send()` doesn't check if `err != nil` before logging.
It led to to logging on every `Send()` call.

Signed-off-by: Ekaterina Lebedeva <ekaterina.lebedeva@yadro.com>
2025-03-07 13:27:07 +00:00
9e31cb249f [] control: Add method to search shards by object
Added method `ListShardsForObject` to ControlService and to
StorageEngine. It returns information about shards storing
object on the node.

Signed-off-by: Ekaterina Lebedeva <ekaterina.lebedeva@yadro.com>
2025-03-07 14:32:01 +03:00
6260d703ce
[] treesvc: Disable service config query
By default, gRPC fetches TXT report while resolving a domain.
0914bba6c5/internal/resolver/dns/dns_resolver.go (L336)

This leads to a hanging dial if DNS is unavailable, even though the host
may be specified in `/etc/hosts` (hello, localhost!).

SDK client for the main API uses these options by default.

Refs 

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-03-06 15:15:31 +03:00
471aeeaff3 [] audit: Fix duplicated request logs
When we do `object put` with audit enabled we get several entries
in logs: with and without object id.

`object put` request is logged in 2 places:
1. `(*auditPutStream) CloseAndRecv()` - when the client closes the
   request stream or when stream gets aborted.
2. `(*auditPutStream) Send()` - when stream was NOT aborted.

`Send()` does error check for `ErrAbortStream` because if there
is any other error - CloseAndRecv will not be called and there
won't be any audit log about failed request.
It led to logging on every object chunck put, even if `err == nil`.

Added check for `err != nil` in `Send()` to fix it.

Signed-off-by: Ekaterina Lebedeva <ekaterina.lebedeva@yadro.com>
2025-03-06 12:36:16 +03:00
4c8f9580a1 [] object: Fix CloseAndRecv for patch streamer
* A client may open stream to server, not send anything and close
  the open stream immediatly. This shouldn't cause a panic;
* Return the error if `s.patcher` is uninitialized. Uninitialized
  patcher cannot be closed, this causes a panic.

Signed-off-by: Airat Arifullin <a.arifullin@yadro.com>
2025-03-05 17:16:10 +03:00
4685afb1dc
[] engine: Validate limiter release in unit tests
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:30 +03:00
eb8b9b2b3b
[] blobovniczatree: Validate limiter release in rebuild unit tests
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:30 +03:00
6c6e463b73
[] shard: Change ops limiter on shard reload
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:29 +03:00
c2d855aedd
[] qos: Return Resource Exhausted error
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:28 +03:00
2162f8e189
[] object: Fix IO tag adjustment for Put/Patch
There was no tag adjustment for CloseAndRecv.

Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:28 +03:00
b9360be1dc
[] blobovniczatree: Use RebuildLimiter
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:27 +03:00
ceff5e1f6a
[] storage: Refactor shard rebuild
Drop redundant interfaces.
Rename fields.

Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:27 +03:00
e0dc3c3d0c
[] shard: Add limiter usage
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 17:25:26 +03:00
0991077cb3 [] engine: Fix data race in evacuation tests
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-28 14:14:12 +00:00
dce269c62e [] services/object: Remove limiting pools for Put operation
Signed-off-by: Aleksey Savchuk <a.savchuk@yadro.com>
2025-02-28 11:08:09 +00:00
9a0507704a [] apemanager: Wrap some errors with InvalidArgument status
Signed-off-by: Airat Arifullin <a.arifullin@yadro.com>
2025-02-27 18:25:09 +00:00
003d568ae2
[] innerring: Relax container homomorphic hashing check
Our initial desire was to prohibit using homomorphic hashing on the
network level because of the resource consumption. However, the ability
to use it, doesn't mean that we must. So only fail validation if
container wants to have homomorphic hashing, but the network prohibits
it.

Signed-off-by: Evgenii Stratonikov <e.stratonikov@yadro.com>
2025-02-27 10:20:08 +03:00
02f3a7f65c
[] writecache: Fix race condition when reporting cache size metrics
There is a race condition when multiple cache operation try to report
the cache size metrics simultaneously. Consider the following example:
- the initial total size of objects stored in the cache size is 2
- worker X deletes an object and reads the cache size, which is 1
- worker Y deletes an object and reads the cache size, which is 0
- worker Y reports the cache size it learnt, which is 0
- worker X reports the cache size it learnt, which is 1

As a result, the observed cache size is 1 (i. e. one object remains
in the cache), which is incorrect because the actual cache size is 0.

To fix this, let's report the metrics periodically in the flush loop.

Signed-off-by: Aleksey Savchuk <a.savchuk@yadro.com>
2025-02-19 17:05:40 +03:00
5d79abe523
[] shard: Add IO tag for rebuild
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:07 +03:00
0671c277db
[] tree: Add IO tag for tree sync requests
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:06 +03:00
92450a76ba
[] shard: Add IO tag to GC
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:06 +03:00
abba5b2089
[] writecache: Add IO tag to flush worker
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:06 +03:00
bfe325e61d
[] policer: Add IO tag to context
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:05 +03:00
acec938b2d
[] qos: Add client grpc interceptors
`qos` client interceptors replace internal IO tags `writecache`,
`policer` and `background` with `internal` IO tag for outcomming RPC.

Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:04 +03:00
170860c14a
[] logger: Add IO tag logging
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:03 +03:00
12da2f8262
[] object: Add IO tag adjustment layer
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:03 +03:00
f6b3f79e89
[] qos: Add qos service to adjust incoming IO tags
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:43:03 +03:00
9729f31e5c
[] grpc: Add QoS interceptors for server and clients
Signed-off-by: Dmitrii Stepanov <d.stepanov@yadro.com>
2025-02-07 15:41:06 +03:00
9b113c3156
[] morph: Add tracing for morph queries to neo-go
Signed-off-by: Alexander Chuprov <a.chuprov@yadro.com>
2025-02-05 16:38:20 +03:00