# 混合HDD/SDD的IO scheduler设置

## 说明

大家都知道[磁盘IO调度器](/cloud-atlas-draft/os/linux/storage/disk/disk_io_scheduler.md)平衡了IO吞吐量和IO响应时间，针对不同的磁盘类型，HDD或SSD，提供了相应的优化调度方法。通常，内核编译时候，选择CFQ作为默认的IO scheduler，适合于HDD设备。如果我们使用SSD设备（如现在主流的笔记本），则可以在Linux启动的时候，通过Grub向内核传递 `elevator=deadline`来更改访问SSD设备的IO scheduler。

但是，在线上的服务器，有些针对特定应用采用的混合磁盘类型的服务器，如采用部分SSD磁盘，部分HDD磁盘，则采用这种一刀切调整IO scheduler就不合理，不论是采用`CFQ`还是`deadline`都无法完全满足要求。

这就要求我们的设置方法能够根据服务器安装的磁盘类型，自动检测并采用对应的IO scheduler。

**`动态调整混合存储的io scheduler`**

## 找出系统所有磁盘设备

```bash
ls /sys/block/ | grep sd
```

> 参考 [Find all storage devices attached to a Linux machine](http://stackoverflow.com/questions/200960/find-all-storage-devices-attached-to-a-linux-machine)。注意：如果要找到所有分区可以查看`/proc/partitions`。注意，块设备有很多，其中磁盘设备以`sd`开头。

## 分辨设备是SSD还是HDD

设备是SSD还是HDD可以通过`cat /sys/block/sda/queue/rotational`内容来区分，其中内容为`1`表示是传统磁盘HDD，如果内容是`0`则表示SSD

> 参考[How to know if a disk is an SSD or an HDD](http://unix.stackexchange.com/questions/65595/how-to-know-if-a-disk-is-an-ssd-or-an-hdd)

## 设置SSD设备io scheduler

可以`on fly`方式动态修改SSD设备的io scheduler

```bash
echo deadline > /sys/block/sda/queue/scheduler
```

## 结合脚本命令设置磁盘io scheduler

* 扫描所有磁盘检查是否是SSD（这个命令也就是先验证一遍是否是SSD，为后面准备批量调整）

```bash
for i in `ls /sys/block/ | grep sd`;do ( if [[ `cat /sys/block/$i/queue/rotational` == 0 ]]; then echo "$i ok";else echo "$i ng";fi );done
```

可以看到输出内容符合预期

```bash
sda ok
sdb ng
sdc ng
sdd ng
sde ng
sdf ng
sdg ng
sdh ng
sdi ng
sdj ng
sdk ok
sdl ok
```

* 现在我们可以一次性调整好所有符合要求的SSD的io scheduler为`deadline`，HDD的io scheduler为`cfq` (需要使用`root`权限调整)

```bash
for i in `ls /sys/block/ | grep sd`;do ( if [[ `cat /sys/block/$i/queue/rotational` == 0 ]]; then echo deadline > /sys/block/$i/queue/scheduler;else echo cfq > /sys/block/$i/queue/scheduler;fi );done
```

* 验证调整是否成功

```bash
for i in `ls /sys/block/ | grep sd`;do ( echo $i; cat /sys/block/$i/queue/scheduler );done
```

检查结果符合预期

```bash
sda
noop anticipatory [deadline] cfq 
sdb
noop anticipatory deadline [cfq] 
sdc
noop anticipatory deadline [cfq] 
sdd
noop anticipatory deadline [cfq] 
sde
noop anticipatory deadline [cfq] 
sdf
noop anticipatory deadline [cfq] 
sdg
noop anticipatory deadline [cfq] 
sdh
noop anticipatory deadline [cfq] 
sdi
noop anticipatory deadline [cfq] 
sdj
noop anticipatory deadline [cfq] 
sdk
noop anticipatory [deadline] cfq 
sdl
noop anticipatory [deadline] cfq
```

**`持久化调整混合存储的io scheduler`**

Linux 的 udev rule 可以针对不同的磁盘类型设置io scheduler，方法如下

* 编辑`/etc/udev/rules.d/60-ssd-scheduler.rules`内容如下

```bash
# set deadline scheduler for non-rotating disks
ACTION=="add|change", KERNEL=="sd[a-z]", ATTR{queue/rotational}=="0", ATTR{queue/scheduler}="deadline"
```

* 刷新udev规则

```bash
udevcontrol reload_rules
```

* 以上配置就使得当系统重启或者插拔SSD磁盘时，系统就会自动设置好SSD的io scheduler。可以重启一台NC验证一下

> 参考[Debian - SSDOptimization](https://wiki.debian.org/SSDOptimization#Low-Latency_IO-Scheduler)


---

# Agent Instructions: Querying This Documentation

If you need additional information that is not directly available in this page, you can query the documentation dynamically by asking a question.

Perform an HTTP GET request on the current page URL with the `ask` query parameter:

```
GET https://huataihuang.gitbook.io/cloud-atlas-draft/os/linux/storage/disk/ssd_hdd_mix_io_scheduler.md?ask=<question>
```

The question should be specific, self-contained, and written in natural language.
The response will contain a direct answer to the question and relevant excerpts and sources from the documentation.

Use this mechanism when the answer is not explicitly present in the current page, you need clarification or additional context, or you want to retrieve related documentation sections.