[feat]: change erasure coding default block size from 10MiB to 1MiB (#11721)

major performance improvements in range GETs to avoid large
read amplification when ranges are tiny and random

```
-------------------
Operation: GET
Operations: 142014 -> 339421
Duration: 4m50s -> 4m56s
* Average: +139.41% (+1177.3 MiB/s) throughput, +139.11% (+658.4) obj/s
* Fastest: +125.24% (+1207.4 MiB/s) throughput, +132.32% (+612.9) obj/s
* 50% Median: +139.06% (+1175.7 MiB/s) throughput, +133.46% (+660.9) obj/s
* Slowest: +203.40% (+1267.9 MiB/s) throughput, +198.59% (+753.5) obj/s
```

TTFB from 10MiB BlockSize
```
* First Access TTFB: Avg: 81ms, Median: 61ms, Best: 20ms, Worst: 2.056s
```

TTFB from 1MiB BlockSize
```
* First Access TTFB: Avg: 22ms, Median: 21ms, Best: 8ms, Worst: 91ms
```

Full object reads however do see a slight change which won't be
noticeable in real world, so not doing any comparisons

TTFB still had improvements with full object reads with 1MiB

```
* First Access TTFB: Avg: 68ms, Median: 35ms, Best: 11ms, Worst: 1.16s
```

v/s

TTFB with 10MiB
```
* First Access TTFB: Avg: 388ms, Median: 98ms, Best: 20ms, Worst: 4.156s
```

This change should affect all new uploads, previous uploads should
continue to work with business as usual. But dramatic improvements can
be seen with these changes.
This commit is contained in:
Harshavardhana 2021-03-06 14:09:34 -08:00 committed by GitHub
parent abce040088
commit 9ccc483df6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
15 changed files with 150 additions and 135 deletions

View File

@ -1,4 +1,4 @@
FROM golang:1.15-alpine as builder FROM golang:1.16-alpine as builder
LABEL maintainer="MinIO Inc <dev@min.io>" LABEL maintainer="MinIO Inc <dev@min.io>"

View File

@ -1,4 +1,4 @@
FROM golang:1.15-alpine as builder FROM golang:1.16-alpine as builder
LABEL maintainer="MinIO Inc <dev@min.io>" LABEL maintainer="MinIO Inc <dev@min.io>"

100
README.md
View File

@ -5,32 +5,32 @@
MinIO is a High Performance Object Storage released under Apache License v2.0. It is API compatible with Amazon S3 cloud storage service. Use MinIO to build high performance infrastructure for machine learning, analytics and application data workloads. MinIO is a High Performance Object Storage released under Apache License v2.0. It is API compatible with Amazon S3 cloud storage service. Use MinIO to build high performance infrastructure for machine learning, analytics and application data workloads.
This README provides quickstart instructions on running MinIO on baremetal hardware, including Docker-based installations. For Kubernetes environments, This README provides quickstart instructions on running MinIO on baremetal hardware, including Docker-based installations. For Kubernetes environments,
use the [MinIO Kubernetes Operator](https://github.com/minio/operator/blob/master/README.md). use the [MinIO Kubernetes Operator](https://github.com/minio/operator/blob/master/README.md).
# Docker Installation # Docker Installation
Use the following commands to run a standalone MinIO server on a Docker container. Use the following commands to run a standalone MinIO server on a Docker container.
Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication
require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically, require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically,
with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html) with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html)
for more complete documentation. for more complete documentation.
## Stable ## Stable
Run the following command to run the latest stable image of MinIO on a Docker container using an ephemeral data volume: Run the following command to run the latest stable image of MinIO on a Docker container using an ephemeral data volume:
```sh ```sh
docker run -p 9000:9000 minio/minio server /data docker run -p 9000:9000 minio/minio server /data
``` ```
The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded
web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the
root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server. root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.
You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See
[Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers, [Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers,
see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages. see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages.
@ -45,12 +45,12 @@ Run the following command to run the bleeding-edge image of MinIO on a Docker co
docker run -p 9000:9000 minio/minio:edge server /data docker run -p 9000:9000 minio/minio:edge server /data
``` ```
The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded
web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the
root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server. root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.
You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See
[Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers, [Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers,
see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages. see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages.
@ -61,9 +61,9 @@ see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view
Use the following commands to run a standalone MinIO server on macOS. Use the following commands to run a standalone MinIO server on macOS.
Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication
require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically, require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically,
with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html) with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html)
for more complete documentation. for more complete documentation.
## Homebrew (recommended) ## Homebrew (recommended)
@ -82,12 +82,12 @@ brew uninstall minio
brew install minio/stable/minio brew install minio/stable/minio
``` ```
The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded
web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the
root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server. root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.
You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See
[Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers, [Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers,
see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages. see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages.
## Binary Download ## Binary Download
@ -100,12 +100,12 @@ chmod +x minio
./minio server /data ./minio server /data
``` ```
The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded
web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the
root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server. root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.
You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See
[Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers, [Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers,
see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages. see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages.
@ -130,18 +130,18 @@ The following table lists supported architectures. Replace the `wget` URL with t
| 64-bit PowerPC LE (ppc64le) | https://dl.min.io/server/minio/release/linux-ppc64le/minio | | 64-bit PowerPC LE (ppc64le) | https://dl.min.io/server/minio/release/linux-ppc64le/minio |
| IBM Z-Series (S390X) | https://dl.min.io/server/minio/release/linux-s390x/minio | | IBM Z-Series (S390X) | https://dl.min.io/server/minio/release/linux-s390x/minio |
The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded
web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the
root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server. root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.
You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See
[Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers, [Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers,
see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages. see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages.
> NOTE: Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication > NOTE: Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication
require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically, require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically,
with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html) with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html)
for more complete documentation. for more complete documentation.
# Microsoft Windows # Microsoft Windows
@ -158,17 +158,17 @@ Use the following command to run a standalone MinIO server on the Windows host.
minio.exe server D:\ minio.exe server D:\
``` ```
The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded
web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the
root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server. root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.
You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See
[Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers, [Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers,
see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages. see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages.
> NOTE: Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication > NOTE: Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication
require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically, require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically,
with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html) with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html)
for more complete documentation. for more complete documentation.
# FreeBSD # FreeBSD
@ -184,27 +184,27 @@ service minio start
# Install from Source # Install from Source
Use the following commands to compile and run a standalone MinIO server from source. Source installation is only intended for developers and advanced users. If you do not have a working Golang environment, please follow [How to install Golang](https://golang.org/doc/install). Minimum version required is [go1.15](https://golang.org/dl/#stable) Use the following commands to compile and run a standalone MinIO server from source. Source installation is only intended for developers and advanced users. If you do not have a working Golang environment, please follow [How to install Golang](https://golang.org/doc/install). Minimum version required is [go1.16](https://golang.org/dl/#stable)
```sh ```sh
GO111MODULE=on go get github.com/minio/minio GO111MODULE=on go get github.com/minio/minio
``` ```
The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded The MinIO deployment starts using default root credentials `minioadmin:minioadmin`. You can test the deployment using the MinIO Browser, an embedded
web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the web-based object browser built into MinIO Server. Point a web browser running on the host machine to http://127.0.0.1:9000 and log in with the
root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server. root credentials. You can use the Browser to create buckets, upload objects, and browse the contents of the MinIO server.
You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See You can also connect using any S3-compatible tool, such as the MinIO Client `mc` commandline tool. See
[Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers, [Test using MinIO Client `mc`](#test-using-minio-client-mc) for more information on using the `mc` commandline tool. For application developers,
see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages. see https://docs.min.io/docs/ and click **MINIO SDKS** in the navigation to view MinIO SDKs for supported languages.
> NOTE: Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication > NOTE: Standalone MinIO servers are best suited for early development and evaluation. Certain features such as versioning, object locking, and bucket replication
require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically, require distributed deploying MinIO with Erasure Coding. For extended development and production, deploy MinIO with Erasure Coding enabled - specifically,
with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html) with a *minimum* of 4 drives per MinIO server. See [MinIO Erasure Code Quickstart Guide](https://docs.min.io/docs/minio-erasure-code-quickstart-guide.html)
for more complete documentation. for more complete documentation.
MinIO strongly recommends *against* using compiled-from-source MinIO servers for production environments. MinIO strongly recommends *against* using compiled-from-source MinIO servers for production environments.
# Deployment Recommendations # Deployment Recommendations

View File

@ -89,7 +89,7 @@ service minio start
## 使用源码安装 ## 使用源码安装
采用源码安装仅供开发人员和高级用户使用,如果你还没有Golang环境 请参考 [How to install Golang](https://golang.org/doc/install)。最低需要Golang版本为 [go1.15](https://golang.org/dl/#stable) 采用源码安装仅供开发人员和高级用户使用,如果你还没有Golang环境 请参考 [How to install Golang](https://golang.org/doc/install)。最低需要Golang版本为 [go1.16](https://golang.org/dl/#stable)
```sh ```sh
GO111MODULE=on go get github.com/minio/minio GO111MODULE=on go get github.com/minio/minio

View File

@ -21,7 +21,7 @@ _init() {
## Minimum required versions for build dependencies ## Minimum required versions for build dependencies
GIT_VERSION="1.0" GIT_VERSION="1.0"
GO_VERSION="1.13" GO_VERSION="1.16"
OSX_VERSION="10.8" OSX_VERSION="10.8"
KNAME=$(uname -s) KNAME=$(uname -s)
ARCH=$(uname -m) ARCH=$(uname -m)

View File

@ -41,46 +41,46 @@ var erasureDecodeTests = []struct {
algorithm BitrotAlgorithm algorithm BitrotAlgorithm
shouldFail, shouldFailQuorum bool shouldFail, shouldFailQuorum bool
}{ }{
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 0 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 0
{dataBlocks: 3, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false}, // 1 {dataBlocks: 3, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false}, // 1
{dataBlocks: 4, onDisks: 8, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 2 {dataBlocks: 4, onDisks: 8, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 2
{dataBlocks: 5, onDisks: 10, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 1, length: oneMiByte - 1, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 3 {dataBlocks: 5, onDisks: 10, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 1, length: oneMiByte - 1, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 3
{dataBlocks: 6, onDisks: 12, offDisks: 0, blocksize: int64(oneMiByte), data: oneMiByte, offset: oneMiByte, length: 0, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, {dataBlocks: 6, onDisks: 12, offDisks: 0, blocksize: int64(oneMiByte), data: oneMiByte, offset: oneMiByte, length: 0, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false},
// 4 // 4
{dataBlocks: 7, onDisks: 14, offDisks: 0, blocksize: int64(oneMiByte), data: oneMiByte, offset: 3, length: 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 5 {dataBlocks: 7, onDisks: 14, offDisks: 0, blocksize: int64(oneMiByte), data: oneMiByte, offset: 3, length: 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 5
{dataBlocks: 8, onDisks: 16, offDisks: 0, blocksize: int64(oneMiByte), data: oneMiByte, offset: 4, length: 8 * 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 6 {dataBlocks: 8, onDisks: 16, offDisks: 0, blocksize: int64(oneMiByte), data: oneMiByte, offset: 4, length: 8 * 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 6
{dataBlocks: 7, onDisks: 14, offDisks: 7, blocksize: int64(blockSizeV1), data: oneMiByte, offset: oneMiByte, length: 1, algorithm: DefaultBitrotAlgorithm, shouldFail: true, shouldFailQuorum: false}, // 7 {dataBlocks: 7, onDisks: 14, offDisks: 7, blocksize: int64(blockSizeV2), data: oneMiByte, offset: oneMiByte, length: 1, algorithm: DefaultBitrotAlgorithm, shouldFail: true, shouldFailQuorum: false}, // 7
{dataBlocks: 6, onDisks: 12, offDisks: 6, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 8 {dataBlocks: 6, onDisks: 12, offDisks: 6, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 8
{dataBlocks: 5, onDisks: 10, offDisks: 5, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 9 {dataBlocks: 5, onDisks: 10, offDisks: 5, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 9
{dataBlocks: 4, onDisks: 8, offDisks: 4, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false}, // 10 {dataBlocks: 4, onDisks: 8, offDisks: 4, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false}, // 10
{dataBlocks: 3, onDisks: 6, offDisks: 3, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 11 {dataBlocks: 3, onDisks: 6, offDisks: 3, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 11
{dataBlocks: 2, onDisks: 4, offDisks: 2, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 12 {dataBlocks: 2, onDisks: 4, offDisks: 2, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 12
{dataBlocks: 2, onDisks: 4, offDisks: 1, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 13 {dataBlocks: 2, onDisks: 4, offDisks: 1, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 13
{dataBlocks: 3, onDisks: 6, offDisks: 2, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 14 {dataBlocks: 3, onDisks: 6, offDisks: 2, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 14
{dataBlocks: 4, onDisks: 8, offDisks: 3, blocksize: int64(2 * oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 15 {dataBlocks: 4, onDisks: 8, offDisks: 3, blocksize: int64(2 * oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 15
{dataBlocks: 5, onDisks: 10, offDisks: 6, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 16 {dataBlocks: 5, onDisks: 10, offDisks: 6, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 16
{dataBlocks: 5, onDisks: 10, offDisks: 2, blocksize: int64(blockSizeV1), data: 2 * oneMiByte, offset: oneMiByte, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 17 {dataBlocks: 5, onDisks: 10, offDisks: 2, blocksize: int64(blockSizeV2), data: 2 * oneMiByte, offset: oneMiByte, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 17
{dataBlocks: 5, onDisks: 10, offDisks: 1, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 18 {dataBlocks: 5, onDisks: 10, offDisks: 1, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 18
{dataBlocks: 6, onDisks: 12, offDisks: 3, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false}, {dataBlocks: 6, onDisks: 12, offDisks: 3, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false},
// 19 // 19
{dataBlocks: 6, onDisks: 12, offDisks: 7, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 20 {dataBlocks: 6, onDisks: 12, offDisks: 7, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 20
{dataBlocks: 8, onDisks: 16, offDisks: 8, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 21 {dataBlocks: 8, onDisks: 16, offDisks: 8, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 21
{dataBlocks: 8, onDisks: 16, offDisks: 9, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 22 {dataBlocks: 8, onDisks: 16, offDisks: 9, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 22
{dataBlocks: 8, onDisks: 16, offDisks: 7, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 23 {dataBlocks: 8, onDisks: 16, offDisks: 7, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 23
{dataBlocks: 2, onDisks: 4, offDisks: 1, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 24 {dataBlocks: 2, onDisks: 4, offDisks: 1, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 24
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 25 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, length: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 25
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(blockSizeV1) + 1, offset: 0, length: int64(blockSizeV1) + 1, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 26 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(blockSizeV2) + 1, offset: 0, length: int64(blockSizeV2) + 1, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 26
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: 12, length: int64(blockSizeV1) + 17, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 27 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: 12, length: int64(blockSizeV2) + 17, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 27
{dataBlocks: 3, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: 1023, length: int64(blockSizeV1) + 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 28 {dataBlocks: 3, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: 1023, length: int64(blockSizeV2) + 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 28
{dataBlocks: 4, onDisks: 8, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: 11, length: int64(blockSizeV1) + 2*1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 29 {dataBlocks: 4, onDisks: 8, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: 11, length: int64(blockSizeV2) + 2*1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 29
{dataBlocks: 6, onDisks: 12, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: 512, length: int64(blockSizeV1) + 8*1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 30 {dataBlocks: 6, onDisks: 12, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: 512, length: int64(blockSizeV2) + 8*1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 30
{dataBlocks: 8, onDisks: 16, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: int64(blockSizeV1), length: int64(blockSizeV1) - 1, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 31 {dataBlocks: 8, onDisks: 16, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: int64(blockSizeV2), length: int64(blockSizeV2) - 1, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 31
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(oneMiByte), offset: -1, length: 3, algorithm: DefaultBitrotAlgorithm, shouldFail: true, shouldFailQuorum: false}, // 32 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(oneMiByte), offset: -1, length: 3, algorithm: DefaultBitrotAlgorithm, shouldFail: true, shouldFailQuorum: false}, // 32
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(oneMiByte), offset: 1024, length: -1, algorithm: DefaultBitrotAlgorithm, shouldFail: true, shouldFailQuorum: false}, // 33 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(oneMiByte), offset: 1024, length: -1, algorithm: DefaultBitrotAlgorithm, shouldFail: true, shouldFailQuorum: false}, // 33
{dataBlocks: 4, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV1), data: int64(blockSizeV1), offset: 0, length: int64(blockSizeV1), algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 34 {dataBlocks: 4, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV2), data: int64(blockSizeV2), offset: 0, length: int64(blockSizeV2), algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 34
{dataBlocks: 4, onDisks: 6, offDisks: 1, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: 12, length: int64(blockSizeV1) + 17, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 35 {dataBlocks: 4, onDisks: 6, offDisks: 1, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: 12, length: int64(blockSizeV2) + 17, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 35
{dataBlocks: 4, onDisks: 6, offDisks: 3, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: 1023, length: int64(blockSizeV1) + 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 36 {dataBlocks: 4, onDisks: 6, offDisks: 3, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: 1023, length: int64(blockSizeV2) + 1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 36
{dataBlocks: 8, onDisks: 12, offDisks: 4, blocksize: int64(blockSizeV1), data: int64(2 * blockSizeV1), offset: 11, length: int64(blockSizeV1) + 2*1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 37 {dataBlocks: 8, onDisks: 12, offDisks: 4, blocksize: int64(blockSizeV2), data: int64(2 * blockSizeV2), offset: 11, length: int64(blockSizeV2) + 2*1024, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 37
} }
func TestErasureDecode(t *testing.T) { func TestErasureDecode(t *testing.T) {
@ -288,13 +288,13 @@ func TestErasureDecodeRandomOffsetLength(t *testing.T) {
// Benchmarks // Benchmarks
func benchmarkErasureDecode(data, parity, dataDown, parityDown int, size int64, b *testing.B) { func benchmarkErasureDecode(data, parity, dataDown, parityDown int, size int64, b *testing.B) {
setup, err := newErasureTestSetup(data, parity, blockSizeV1) setup, err := newErasureTestSetup(data, parity, blockSizeV2)
if err != nil { if err != nil {
b.Fatalf("failed to create test setup: %v", err) b.Fatalf("failed to create test setup: %v", err)
} }
defer setup.Remove() defer setup.Remove()
disks := setup.disks disks := setup.disks
erasure, err := NewErasure(context.Background(), data, parity, blockSizeV1) erasure, err := NewErasure(context.Background(), data, parity, blockSizeV2)
if err != nil { if err != nil {
b.Fatalf("failed to create ErasureStorage: %v", err) b.Fatalf("failed to create ErasureStorage: %v", err)
} }
@ -308,7 +308,7 @@ func benchmarkErasureDecode(data, parity, dataDown, parityDown int, size int64,
} }
content := make([]byte, size) content := make([]byte, size)
buffer := make([]byte, blockSizeV1, 2*blockSizeV1) buffer := make([]byte, blockSizeV2, 2*blockSizeV2)
_, err = erasure.Encode(context.Background(), bytes.NewReader(content), writers, buffer, erasure.dataBlocks+1) _, err = erasure.Encode(context.Background(), bytes.NewReader(content), writers, buffer, erasure.dataBlocks+1)
closeBitrotWriters(writers) closeBitrotWriters(writers)
if err != nil { if err != nil {

View File

@ -62,26 +62,26 @@ var erasureEncodeTests = []struct {
algorithm BitrotAlgorithm algorithm BitrotAlgorithm
shouldFail, shouldFailQuorum bool shouldFail, shouldFailQuorum bool
}{ }{
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 0 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 0
{dataBlocks: 3, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 1, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false}, // 1 {dataBlocks: 3, onDisks: 6, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 1, algorithm: SHA256, shouldFail: false, shouldFailQuorum: false}, // 1
{dataBlocks: 4, onDisks: 8, offDisks: 2, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 2 {dataBlocks: 4, onDisks: 8, offDisks: 2, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 2
{dataBlocks: 5, onDisks: 10, offDisks: 3, blocksize: int64(blockSizeV1), data: oneMiByte, offset: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 3 {dataBlocks: 5, onDisks: 10, offDisks: 3, blocksize: int64(blockSizeV2), data: oneMiByte, offset: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 3
{dataBlocks: 6, onDisks: 12, offDisks: 4, blocksize: int64(blockSizeV1), data: oneMiByte, offset: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 4 {dataBlocks: 6, onDisks: 12, offDisks: 4, blocksize: int64(blockSizeV2), data: oneMiByte, offset: oneMiByte, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 4
{dataBlocks: 7, onDisks: 14, offDisks: 5, blocksize: int64(blockSizeV1), data: 0, offset: 0, shouldFail: false, algorithm: SHA256, shouldFailQuorum: false}, // 5 {dataBlocks: 7, onDisks: 14, offDisks: 5, blocksize: int64(blockSizeV2), data: 0, offset: 0, shouldFail: false, algorithm: SHA256, shouldFailQuorum: false}, // 5
{dataBlocks: 8, onDisks: 16, offDisks: 7, blocksize: int64(blockSizeV1), data: 0, offset: 0, shouldFail: false, algorithm: DefaultBitrotAlgorithm, shouldFailQuorum: false}, // 6 {dataBlocks: 8, onDisks: 16, offDisks: 7, blocksize: int64(blockSizeV2), data: 0, offset: 0, shouldFail: false, algorithm: DefaultBitrotAlgorithm, shouldFailQuorum: false}, // 6
{dataBlocks: 2, onDisks: 4, offDisks: 2, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: true}, // 7 {dataBlocks: 2, onDisks: 4, offDisks: 2, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: true}, // 7
{dataBlocks: 4, onDisks: 8, offDisks: 4, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, algorithm: SHA256, shouldFail: false, shouldFailQuorum: true}, // 8 {dataBlocks: 4, onDisks: 8, offDisks: 4, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, algorithm: SHA256, shouldFail: false, shouldFailQuorum: true}, // 8
{dataBlocks: 7, onDisks: 14, offDisks: 7, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 9 {dataBlocks: 7, onDisks: 14, offDisks: 7, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 9
{dataBlocks: 8, onDisks: 16, offDisks: 8, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 10 {dataBlocks: 8, onDisks: 16, offDisks: 8, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 10
{dataBlocks: 5, onDisks: 10, offDisks: 3, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 11 {dataBlocks: 5, onDisks: 10, offDisks: 3, blocksize: int64(oneMiByte), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 11
{dataBlocks: 3, onDisks: 6, offDisks: 1, blocksize: int64(blockSizeV1), data: oneMiByte, offset: oneMiByte / 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 12 {dataBlocks: 3, onDisks: 6, offDisks: 1, blocksize: int64(blockSizeV2), data: oneMiByte, offset: oneMiByte / 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 12
{dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(oneMiByte / 2), data: oneMiByte, offset: oneMiByte/2 + 1, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 13 {dataBlocks: 2, onDisks: 4, offDisks: 0, blocksize: int64(oneMiByte / 2), data: oneMiByte, offset: oneMiByte/2 + 1, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 13
{dataBlocks: 4, onDisks: 8, offDisks: 0, blocksize: int64(oneMiByte - 1), data: oneMiByte, offset: oneMiByte - 1, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 14 {dataBlocks: 4, onDisks: 8, offDisks: 0, blocksize: int64(oneMiByte - 1), data: oneMiByte, offset: oneMiByte - 1, algorithm: BLAKE2b512, shouldFail: false, shouldFailQuorum: false}, // 14
{dataBlocks: 8, onDisks: 12, offDisks: 2, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 15 {dataBlocks: 8, onDisks: 12, offDisks: 2, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 15
{dataBlocks: 8, onDisks: 10, offDisks: 1, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 16 {dataBlocks: 8, onDisks: 10, offDisks: 1, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 16
{dataBlocks: 10, onDisks: 14, offDisks: 0, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 17, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 17 {dataBlocks: 10, onDisks: 14, offDisks: 0, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 17, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 17
{dataBlocks: 2, onDisks: 6, offDisks: 2, blocksize: int64(oneMiByte), data: oneMiByte, offset: oneMiByte / 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 18 {dataBlocks: 2, onDisks: 6, offDisks: 2, blocksize: int64(oneMiByte), data: oneMiByte, offset: oneMiByte / 2, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: false}, // 18
{dataBlocks: 10, onDisks: 16, offDisks: 8, blocksize: int64(blockSizeV1), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 19 {dataBlocks: 10, onDisks: 16, offDisks: 8, blocksize: int64(blockSizeV2), data: oneMiByte, offset: 0, algorithm: DefaultBitrotAlgorithm, shouldFail: false, shouldFailQuorum: true}, // 19
} }
func TestErasureEncode(t *testing.T) { func TestErasureEncode(t *testing.T) {
@ -166,17 +166,17 @@ func TestErasureEncode(t *testing.T) {
// Benchmarks // Benchmarks
func benchmarkErasureEncode(data, parity, dataDown, parityDown int, size int64, b *testing.B) { func benchmarkErasureEncode(data, parity, dataDown, parityDown int, size int64, b *testing.B) {
setup, err := newErasureTestSetup(data, parity, blockSizeV1) setup, err := newErasureTestSetup(data, parity, blockSizeV2)
if err != nil { if err != nil {
b.Fatalf("failed to create test setup: %v", err) b.Fatalf("failed to create test setup: %v", err)
} }
defer setup.Remove() defer setup.Remove()
erasure, err := NewErasure(context.Background(), data, parity, blockSizeV1) erasure, err := NewErasure(context.Background(), data, parity, blockSizeV2)
if err != nil { if err != nil {
b.Fatalf("failed to create ErasureStorage: %v", err) b.Fatalf("failed to create ErasureStorage: %v", err)
} }
disks := setup.disks disks := setup.disks
buffer := make([]byte, blockSizeV1, 2*blockSizeV1) buffer := make([]byte, blockSizeV2, 2*blockSizeV2)
content := make([]byte, size) content := make([]byte, size)
for i := 0; i < dataDown; i++ { for i := 0; i < dataDown; i++ {

View File

@ -39,26 +39,26 @@ var erasureHealTests = []struct {
algorithm BitrotAlgorithm algorithm BitrotAlgorithm
shouldFail bool shouldFail bool
}{ }{
{dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: SHA256, shouldFail: false}, // 0 {dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: SHA256, shouldFail: false}, // 0
{dataBlocks: 3, disks: 6, offDisks: 2, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 1 {dataBlocks: 3, disks: 6, offDisks: 2, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 1
{dataBlocks: 4, disks: 8, offDisks: 2, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 2 {dataBlocks: 4, disks: 8, offDisks: 2, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 2
{dataBlocks: 5, disks: 10, offDisks: 3, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 3 {dataBlocks: 5, disks: 10, offDisks: 3, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 3
{dataBlocks: 6, disks: 12, offDisks: 2, badDisks: 3, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: SHA256, shouldFail: false}, // 4 {dataBlocks: 6, disks: 12, offDisks: 2, badDisks: 3, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: SHA256, shouldFail: false}, // 4
{dataBlocks: 7, disks: 14, offDisks: 4, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 5 {dataBlocks: 7, disks: 14, offDisks: 4, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 5
{dataBlocks: 8, disks: 16, offDisks: 6, badDisks: 1, badStaleDisks: 1, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 6 {dataBlocks: 8, disks: 16, offDisks: 6, badDisks: 1, badStaleDisks: 1, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 6
{dataBlocks: 7, disks: 14, offDisks: 2, badDisks: 3, badStaleDisks: 0, blocksize: int64(oneMiByte / 2), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 7 {dataBlocks: 7, disks: 14, offDisks: 2, badDisks: 3, badStaleDisks: 0, blocksize: int64(oneMiByte / 2), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 7
{dataBlocks: 6, disks: 12, offDisks: 1, badDisks: 0, badStaleDisks: 1, blocksize: int64(oneMiByte - 1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 8 {dataBlocks: 6, disks: 12, offDisks: 1, badDisks: 0, badStaleDisks: 1, blocksize: int64(oneMiByte - 1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 8
{dataBlocks: 5, disks: 10, offDisks: 3, badDisks: 0, badStaleDisks: 3, blocksize: int64(oneMiByte / 2), size: oneMiByte, algorithm: SHA256, shouldFail: true}, // 9 {dataBlocks: 5, disks: 10, offDisks: 3, badDisks: 0, badStaleDisks: 3, blocksize: int64(oneMiByte / 2), size: oneMiByte, algorithm: SHA256, shouldFail: true}, // 9
{dataBlocks: 4, disks: 8, offDisks: 1, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 10 {dataBlocks: 4, disks: 8, offDisks: 1, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 10
{dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 1, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 11 {dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 1, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 11
{dataBlocks: 6, disks: 12, offDisks: 8, badDisks: 3, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 12 {dataBlocks: 6, disks: 12, offDisks: 8, badDisks: 3, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 12
{dataBlocks: 7, disks: 14, offDisks: 3, badDisks: 4, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 13 {dataBlocks: 7, disks: 14, offDisks: 3, badDisks: 4, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 13
{dataBlocks: 7, disks: 14, offDisks: 6, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 14 {dataBlocks: 7, disks: 14, offDisks: 6, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 14
{dataBlocks: 8, disks: 16, offDisks: 4, badDisks: 5, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 15 {dataBlocks: 8, disks: 16, offDisks: 4, badDisks: 5, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: true}, // 15
{dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 16 {dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 16
{dataBlocks: 12, disks: 16, offDisks: 2, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 17 {dataBlocks: 12, disks: 16, offDisks: 2, badDisks: 1, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: DefaultBitrotAlgorithm, shouldFail: false}, // 17
{dataBlocks: 6, disks: 8, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 18 {dataBlocks: 6, disks: 8, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte, algorithm: BLAKE2b512, shouldFail: false}, // 18
{dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV1), size: oneMiByte * 64, algorithm: SHA256, shouldFail: false}, // 19 {dataBlocks: 2, disks: 4, offDisks: 1, badDisks: 0, badStaleDisks: 0, blocksize: int64(blockSizeV2), size: oneMiByte * 64, algorithm: SHA256, shouldFail: false}, // 19
} }
func TestErasureHeal(t *testing.T) { func TestErasureHeal(t *testing.T) {

View File

@ -29,6 +29,7 @@ import (
"time" "time"
"github.com/dchest/siphash" "github.com/dchest/siphash"
"github.com/dustin/go-humanize"
"github.com/google/uuid" "github.com/google/uuid"
"github.com/minio/minio-go/v7/pkg/set" "github.com/minio/minio-go/v7/pkg/set"
"github.com/minio/minio-go/v7/pkg/tags" "github.com/minio/minio-go/v7/pkg/tags"
@ -359,14 +360,12 @@ func newErasureSets(ctx context.Context, endpoints Endpoints, storageDisks []Sto
mutex := newNSLock(globalIsDistErasure) mutex := newNSLock(globalIsDistErasure)
// Number of buffers, max 2GB. // Number of buffers, max 2GB
n := setCount * setDriveCount n := (2 * humanize.GiByte) / (blockSizeV2 * 2)
if n > 100 {
n = 100
}
// Initialize byte pool once for all sets, bpool size is set to // Initialize byte pool once for all sets, bpool size is set to
// setCount * setDriveCount with each memory upto blockSizeV1. // setCount * setDriveCount with each memory upto blockSizeV2.
bp := bpool.NewBytePoolCap(n, blockSizeV1, blockSizeV1*2) bp := bpool.NewBytePoolCap(n, blockSizeV2, blockSizeV2*2)
for i := 0; i < setCount; i++ { for i := 0; i < setCount; i++ {
s.erasureDisks[i] = make([]StorageAPI, setDriveCount) s.erasureDisks[i] = make([]StorageAPI, setDriveCount)

View File

@ -52,7 +52,7 @@ func TestErasureEncodeDecode(t *testing.T) {
buffer := make([]byte, len(data), 2*len(data)) buffer := make([]byte, len(data), 2*len(data))
copy(buffer, data) copy(buffer, data)
erasure, err := NewErasure(context.Background(), test.dataBlocks, test.parityBlocks, blockSizeV1) erasure, err := NewErasure(context.Background(), test.dataBlocks, test.parityBlocks, blockSizeV2)
if err != nil { if err != nil {
t.Fatalf("Test %d: failed to create erasure: %v", i, err) t.Fatalf("Test %d: failed to create erasure: %v", i, err)
} }

View File

@ -55,13 +55,14 @@ func (t *apiConfig) init(cfg api.Config, setDriveCounts []int) {
stats, err := sys.GetStats() stats, err := sys.GetStats()
if err != nil { if err != nil {
logger.LogIf(GlobalContext, err) logger.LogIf(GlobalContext, err)
// Default to 16 GiB, not critical. // Default to 8 GiB, not critical.
stats.TotalRAM = 16 << 30 stats.TotalRAM = 8 << 30
} }
// max requests per node is calculated as // max requests per node is calculated as
// total_ram / ram_per_request // total_ram / ram_per_request
// ram_per_request is (2MiB+128KiB) * driveCount + 2 * 10MiB (default erasure block size) // ram_per_request is (2MiB+128KiB) * driveCount \
apiRequestsMaxPerNode = int(stats.TotalRAM / uint64(t.totalDriveCount*(blockSizeLarge+blockSizeSmall)+blockSizeV1*2)) // + 2 * 10MiB (default erasure block size v1) + 2 * 1MiB (default erasure block size v2)
apiRequestsMaxPerNode = int(stats.TotalRAM / uint64(t.totalDriveCount*(blockSizeLarge+blockSizeSmall)+int(blockSizeV1*2+blockSizeV2*2)))
} else { } else {
apiRequestsMaxPerNode = cfg.RequestsMax apiRequestsMaxPerNode = cfg.RequestsMax
if len(globalEndpoints.Hostnames()) > 0 { if len(globalEndpoints.Hostnames()) > 0 {

View File

@ -29,8 +29,17 @@ import (
const ( const (
// Block size used for all internal operations version 1. // Block size used for all internal operations version 1.
// TLDR..
// Not used anymore xl.meta captures the right blockSize
// so blockSizeV2 should be used for all future purposes.
// this value is kept here to calculate the max API
// requests based on RAM size for existing content.
blockSizeV1 = 10 * humanize.MiByte blockSizeV1 = 10 * humanize.MiByte
// Block size used in erasure coding version 2.
blockSizeV2 = 1 * humanize.MiByte
// Buckets meta prefix. // Buckets meta prefix.
bucketMetaPrefix = "buckets" bucketMetaPrefix = "buckets"

View File

@ -193,7 +193,7 @@ func newFileInfo(object string, dataBlocks, parityBlocks int) (fi FileInfo) {
Algorithm: erasureAlgorithm, Algorithm: erasureAlgorithm,
DataBlocks: dataBlocks, DataBlocks: dataBlocks,
ParityBlocks: parityBlocks, ParityBlocks: parityBlocks,
BlockSize: blockSizeV1, BlockSize: blockSizeV2,
Distribution: hashOrder(object, dataBlocks+parityBlocks), Distribution: hashOrder(object, dataBlocks+parityBlocks),
} }
return fi return fi

View File

@ -451,6 +451,8 @@ func newInternodeHTTPTransport(tlsConfig *tls.Config, dialTimeout time.Duration)
Proxy: http.ProxyFromEnvironment, Proxy: http.ProxyFromEnvironment,
DialContext: xhttp.DialContextWithDNSCache(globalDNSCache, xhttp.NewInternodeDialContext(dialTimeout)), DialContext: xhttp.DialContextWithDNSCache(globalDNSCache, xhttp.NewInternodeDialContext(dialTimeout)),
MaxIdleConnsPerHost: 1024, MaxIdleConnsPerHost: 1024,
WriteBufferSize: 32 << 10, // 32KiB moving up from 4KiB default
ReadBufferSize: 32 << 10, // 32KiB moving up from 4KiB default
IdleConnTimeout: 15 * time.Second, IdleConnTimeout: 15 * time.Second,
ResponseHeaderTimeout: 3 * time.Minute, // Set conservative timeouts for MinIO internode. ResponseHeaderTimeout: 3 * time.Minute, // Set conservative timeouts for MinIO internode.
TLSHandshakeTimeout: 15 * time.Second, TLSHandshakeTimeout: 15 * time.Second,
@ -496,6 +498,8 @@ func newCustomHTTPProxyTransport(tlsConfig *tls.Config, dialTimeout time.Duratio
Proxy: http.ProxyFromEnvironment, Proxy: http.ProxyFromEnvironment,
DialContext: xhttp.DialContextWithDNSCache(globalDNSCache, xhttp.NewInternodeDialContext(dialTimeout)), DialContext: xhttp.DialContextWithDNSCache(globalDNSCache, xhttp.NewInternodeDialContext(dialTimeout)),
MaxIdleConnsPerHost: 1024, MaxIdleConnsPerHost: 1024,
WriteBufferSize: 16 << 10, // 16KiB moving up from 4KiB default
ReadBufferSize: 16 << 10, // 16KiB moving up from 4KiB default
IdleConnTimeout: 15 * time.Second, IdleConnTimeout: 15 * time.Second,
ResponseHeaderTimeout: 30 * time.Minute, // Set larger timeouts for proxied requests. ResponseHeaderTimeout: 30 * time.Minute, // Set larger timeouts for proxied requests.
TLSHandshakeTimeout: 10 * time.Second, TLSHandshakeTimeout: 10 * time.Second,
@ -558,6 +562,8 @@ func newCustomHTTPTransport(tlsConfig *tls.Config, dialTimeout time.Duration) fu
Proxy: http.ProxyFromEnvironment, Proxy: http.ProxyFromEnvironment,
DialContext: xhttp.DialContextWithDNSCache(globalDNSCache, xhttp.NewInternodeDialContext(dialTimeout)), DialContext: xhttp.DialContextWithDNSCache(globalDNSCache, xhttp.NewInternodeDialContext(dialTimeout)),
MaxIdleConnsPerHost: 1024, MaxIdleConnsPerHost: 1024,
WriteBufferSize: 16 << 10, // 16KiB moving up from 4KiB default
ReadBufferSize: 16 << 10, // 16KiB moving up from 4KiB default
IdleConnTimeout: 15 * time.Second, IdleConnTimeout: 15 * time.Second,
ResponseHeaderTimeout: 3 * time.Minute, // Set conservative timeouts for MinIO internode. ResponseHeaderTimeout: 3 * time.Minute, // Set conservative timeouts for MinIO internode.
TLSHandshakeTimeout: 10 * time.Second, TLSHandshakeTimeout: 10 * time.Second,

2
go.mod
View File

@ -1,6 +1,6 @@
module github.com/minio/minio module github.com/minio/minio
go 1.15 go 1.16
require ( require (
cloud.google.com/go v0.39.0 cloud.google.com/go v0.39.0