Helyim

seaweedfs 使用纯 Rust 实现。

特性

附加特性

• 可选择无副本或不同副本级别，对机架和数据中心有感知。
• 自动压缩，在删除或更新后回收磁盘空间。
• 自动主服务器故障转移 - 无单点故障（SPOF）。
• 纠删码 用于温暖存储，Rack-Aware 10.4 纠删码降低存储成本。

用法

默认情况下，主节点运行在端口 9333 上，卷节点运行在端口 8080 上。让我们启动一个主节点和一个卷节点，端口为 8080。理想情况下，它们应该从不同的机器上启动。我们将使用 localhost 作为示例。

Helyim 使用 HTTP REST 操作进行读取、写入和删除。响应为 JSON 或 JSONP 格式。

1. 启动主服务器

cargo run --bin helyim master

2. 启动卷服务器

cargo run --bin helyim volume --port 8080 --folders ./vdata:70 --folders ./v1data:10

3. 写文件

要上传文件：首先，向 /dir/assign 发送 HTTP POST、PUT 或 GET 请求以获取 fid 和卷服务器 URL

curl https://127.0.0.1:9333/dir/assign
{"fid":"6,16b7578a5","url":"127.0.0.1:8080","public_url":"127.0.0.1:8080","count":1,"error":""}

其次，要存储文件内容，从响应中发送 HTTP 多部分 POST 请求到 url + '/' + fid

curl -F file=@./sun.jpg http://127.0.0.1:8080/6,16b7578a5
{"name":"sun.jpg","size":1675569,"error":""}

要更新，发送另一个带有更新文件内容的 POST 请求。

要删除，向同一 url + '/' + fid URL 发送 HTTP DELETE 请求

curl -X DELETE http://127.0.0.1:8080/6,16b7578a5

主服务器故障转移

在初始化 Raft 集群时，启动 Leader 和 Follower 实例时需要指定相同的节点序列。

您可以通过访问 https://127.0.0.1:9333/cluster/status 来查看集群状态。

# start master1
cargo run --release --bin helyim -- master --ip 127.0.0.1 --port 9333 \
      --peers 127.0.0.1:9333 \
      --peers 127.0.0.1:9335 \
      --peers 127.0.0.1:9337
      
      
# start master2
cargo run --release --bin helyim -- master --ip 127.0.0.1 --port 9335 \
      --peers 127.0.0.1:9333 \
      --peers 127.0.0.1:9335 \
      --peers 127.0.0.1:9337
      
# start master3
cargo run --release --bin helyim -- master --ip 127.0.0.1 --port 9337 \
      --peers 127.0.0.1:9333 \
      --peers 127.0.0.1:9335 \
      --peers 127.0.0.1:9337

基准测试

在联想 IdeaPad Pro 16 (2023) 笔记本电脑上的结果，带有 SSD，CPU：14 英特尔酷睿 i9 5.4GHz。

似乎比 seaweedfs 慢，尤其是在读取方面。

➜ ./weed benchmark -server=localhost:9333
This is SeaweedFS version 0.76 linux amd64

------------ Writing Benchmark ----------
Completed 15199 of 1048576 requests, 1.4% 15198.1/s 15.3MB/s
Completed 31887 of 1048576 requests, 3.0% 16687.9/s 16.8MB/s
Completed 48439 of 1048576 requests, 4.6% 16551.6/s 16.7MB/s
...
Completed 994044 of 1048576 requests, 94.8% 16645.2/s 16.8MB/s
Completed 1010800 of 1048576 requests, 96.4% 16755.8/s 16.9MB/s
Completed 1027412 of 1048576 requests, 98.0% 16612.2/s 16.7MB/s
Completed 1044319 of 1048576 requests, 99.6% 16907.0/s 17.0MB/s

Concurrency Level:      16
Time taken for tests:   63.249 seconds
Complete requests:      1048576
Failed requests:        0
Total transferred:      1106759553 bytes
Requests per second:    16578.50 [#/sec]
Transfer rate:          17088.29 [Kbytes/sec]

Connection Times (ms)
              min      avg        max      std
Total:        0.1      0.9       29.8      0.4

Percentage of the requests served within a certain time (ms)
   50%      0.9 ms
   66%      1.0 ms
   75%      1.1 ms
   90%      1.3 ms
   95%      1.5 ms
   98%      1.7 ms
   99%      1.8 ms
  100%     29.8 ms

------------ Randomly Reading Benchmark ----------
Completed 89963 of 1048576 requests, 8.6% 89957.6/s 90.5MB/s
Completed 187560 of 1048576 requests, 17.9% 97597.1/s 98.2MB/s
Completed 283486 of 1048576 requests, 27.0% 95925.8/s 96.6MB/s
Completed 382035 of 1048576 requests, 36.4% 98549.4/s 99.2MB/s
Completed 480649 of 1048576 requests, 45.8% 98613.9/s 99.3MB/s
Completed 583585 of 1048576 requests, 55.7% 102933.7/s 103.6MB/s
Completed 683954 of 1048576 requests, 65.2% 100370.9/s 101.0MB/s
Completed 782522 of 1048576 requests, 74.6% 98567.9/s 99.2MB/s
Completed 883504 of 1048576 requests, 84.3% 100982.7/s 101.7MB/s
Completed 987320 of 1048576 requests, 94.2% 103814.3/s 104.5MB/s

Concurrency Level:      16
Time taken for tests:   10.600 seconds
Complete requests:      1048576
Failed requests:        0
Total transferred:      1106777459 bytes
Requests per second:    98925.73 [#/sec]
Transfer rate:          101969.36 [Kbytes/sec]

Connection Times (ms)
              min      avg        max      std
Total:        0.0      0.1       2.3      0.1

Percentage of the requests served within a certain time (ms)
   50%      0.1 ms
   95%      0.2 ms
   98%      0.4 ms
  100%      2.3 ms

致谢

• seaweedfs - SeaweedFS 是一个快速分布式存储系统，用于 blob、对象、文件和数据湖，适用于数十亿个文件！Blob 存储具有 O(1) 磁盘查找，云分层。Filer 支持云驱动器、跨数据中心活动-活动复制、Kubernetes、POSIX FUSE 挂载、S3 API、S3 网关、Hadoop、WebDAV、加密、纠删码。
• zergling - seaweedFS 用 Rust 重新实现。