Datera LIO Linux SCSI Target 管理手册

需积分: 13 95 浏览量更新于2024-07-20 收藏 2.7MB PDF 举报

"LIO Admin Manual (LIO管理手册) 是一份技术文档，专注于LIO Linux SCSI Target的管理，由Datera, Inc.编写。这份手册适用于系统管理员，旨在提供全面的英文参考资料，帮助用户理解和操作LIO Linux SCSI Target。手册版本为4.0，印刷日期为2015年8月6日。" LIO Linux SCSI Target是开源的存储区域网络（SAN）解决方案，它允许Linux系统作为一个目标设备，支持iSCSI、光纤通道（FC）和其他协议的存储服务。LIO（Lightweight Initiator/Target Offload）项目旨在提供一种轻量级的方法来实现这些协议，使得Linux服务器能够作为存储控制器，为网络上的其他设备提供块级存储访问。在这份LIO管理手册中，读者可以期待找到以下关键知识点： 1. **LIO基础概念**：解释了LIO的基本架构、工作原理以及它如何与Linux内核交互，帮助理解LIO如何处理I/O请求。 2. **安装与配置**：详细指导如何在Linux系统上安装LIO模块，设置所需的内核模块，并配置相应的存储接口和逻辑单元号（LUNs）。 3. **设备管理**：涵盖如何添加、删除和管理物理磁盘、虚拟磁盘（如RAID阵列）以及其他存储设备，以用于LIO服务。 4. **网络配置**：包括iSCSI和FC网络配置，如启动器（initiators）发现、会话建立和连接管理。 5. **安全特性**：讨论了如何实施安全策略，如CHAP身份验证，以保护存储数据和网络通信的安全。 6. **性能优化**：提供关于如何调整LIO设置以优化I/O性能的建议和最佳实践。 7. **故障排查**：包含故障诊断和问题解决的指南，帮助管理员在遇到问题时快速定位并修复。 8. **更新与维护**：说明如何更新LIO组件，以及如何定期检查和保持系统健康。 9. **许可证和法律条款**：手册中包含了版权信息和使用条款，强调未经Datera, Inc.的明确许可，不得复制或分发此文档。 10. **商标与版本变更**：列出了相关的商标信息，并指出文档内容可能会随时更新，无需提前通知，这意味着读者应定期查阅最新版本以获取最准确的信息。 LIO作为开源存储解决方案，它的灵活性和可扩展性使其在各种企业环境中受到欢迎。通过深入阅读这份管理手册，管理员可以有效地管理和利用LIO Linux SCSI Target，为他们的数据中心提供高效、可靠的存储服务。

LIO Admin Manual

When the array and all drive members to be used for the array extension have been identified, and only

then, perform the online capacity expansion or RAID level migration:

# arcconf MODIFY <controller#> FROM <DRIVE#> TO MAX <raid-level> <channel#>

<port#> \ [<channel#> <port#> …]

3.5.2 LSI

Display the logical drive information on a RAID controller:

# MegaCli –LDInfo–L[<enclosure#:slot#>, …] -a<controller#>

List the available disk drive slots and pair down the information a bit:

# MegaCli –PDList –a<controller#> | egrep ‘Adapter|Enclosure|Slot|Inquiry’

Identify the slot numbers of the drive(s):

# MegaCli –PDLocate –PhysDrv[<enclosure#:slot#>, …] -a<controller#>

will help to identify the drive in the chassis by blinking the corresponding activity LED.

When the array and all drive members to be used for the array extension have been identified, and only

then, perform the online capacity expansion or RAID level migration:

# MegaCli –LDRecon -Start -r<raid-level>[Add|Rmv PhysDrv[<enclosure#:slot#,

…]] -L<drive#> –a<controller#>

For instance, perform an online capacity expansion or RAID level migration with the physical drive(s) in

enclosure 0, slot 4 on logical drive 1 in RAID controller 0 as setup above:

# MegaCli –LDRecon -Start -r5[Add PhysDrv[0:4]] -L1 –a0

3.6 Failure Management

A soft error doesn’t necessarily indicate a hard drive failure. For instance, after years of operation, some

drives may occasionally not respond fast enough, so the RAID controller flags them as failed, despite the

fact that they are still fully operational. In such cases, a rescan of the SCSI bus may re-add drives back to

the array without the need of any drive replacements.

However, soft errors are a frequent characteristic of aging devices, and in many cases are indicative of

subsequent hard drive failures, so Datera highly recommends replacing drives that exhibit soft errors.

In any event, failed drives must promptly be replaced with a spare drive. Depending on the underlying

RAID level, a degraded RAID set might not be fault tolerant during RAID reconstruction (e.g., RAID5).

3.6.1 Adaptec

Rescan the SCSI bus:

# arcconf RESCAN <controller#>

A drive failure condition can be identified as follows:

# arcconf GETCONFIG 1 PD | egrep "Device #|State\>|Reported Location|Reported

Channel|S.M.A.R.T. Warnings"

After the failed drive has been replaced with a standby, add a new dedicated hot spare drive:

# arcconf SETSTATE <controller#> DEVICE <channel#> <port#> HSP LOGICALDRIVE

<drive#>

LIO Admin Manual

3.6.2 LSI

A drive failure condition can be identified as follows:

|Error|Failure|Event’

After the failed drive has been replaced with a standby, add a new dedicated hot spare drive:

# MegaCli –PDHSP –Set [–Dedicated –Array<drive-group#> –PhysDrv[<enclosure#:slot#,

…] -a<controller#>

For instance, add a new hot spare in enclosure 0, drive 6 to drive group 0 on controller 0:

# MegaCli –PDHSP –Set –Dedicated –Array0 –PhysDrv[0:6] –a0

3.7 Best Practices

3.7.1 RAID Construction Considerations

Datera recommends against configuring SSDs in RAID10 sets, as that involves a significant loss of

available net storage capacity (which is rather expensive with SSDs) and endurance, without much

performance gain. In practice, the performance gain (IOPS) of SSD RAID10 configurations is typically less

than 15%, at the expense of a loss of 50% of net storage capacity.

Instead, Datera recommends using RAID5 with hot spare drives for SSD based volume groups. The

performance loss of RAID5 vs. RAID10 in terms of latency and IOPS is typically less than 15%, and

considering the much higher performance levels of SSDs doesn’t really matter in practice.

For SSD speed and endurance, Datera recommends using a stripe size of either 64kB or 128kB,

depending on the SSD size, in conjunction with RAID5.

As a further optimization, Datera recommends keeping the SSDs net storage capacity below the 80%

utilization mark, which results in in more robust long-term performance, especially in comparison with

RAID10 with less disk space.

For HDD arrays with an emphasis on high-performance and fault tolerance, RAID10 typically provides

the best tradeoffs between performance, reliability and price in a single-shelf environment. Depending

on the block size of filesystem, Datera recommends using a stripe size that equals 8x to 16x of the

defined block size.

Example: For a block size of 4KiB with an ext4 filesystem, Datera recommends a stripe size of 32kB to

64kB as “guidance.” If the filesystem typically uses larger files, larger stripe sizes will result in higher

sustained performance. This is because the block pre-allocation on the drive can be done in larger

chunks, and therefore reduce the amount of head seeks operations necessary to be performed on the

HDDs.

LIO Admin Manual

4 LVM Setup

4.1 Overview

For ease of configuration and modularity, while maintaining scalability, Datera recommends using the

Logical Volume Manager (LVM).

4.2 Initializing Disks or Disk Partitions

Before using a drive or a drive partition as a physical volume, it must be properly initialized:

# pvcreate <device>

Warning: Not initializing the LVM device with pvcreate can cause subtle integrity errors on

the backstores that later result in system instabilities and crashes.

For instance, fast initialization of RAID devices from the LSI RAID BIOS creates seemingly

valid RAID devices that can cause system crashes during operation under load.

In order to use a whole drive as the physical volume, specify a raw disk device for <device>, e.g.,

“/dev/sdb.” This creates a volume group descriptor at the beginning of the drive “/dev/sdb.”

On the other hand, to use a logical partition on a disk device, specify the corresponding primary or

extended partition for <device>, e.g., “/dev/sdb1.” This creates a volume group descriptor at the

beginning of the logical partition “/dev/sdb1.”

Note: LVM works fine with whole disk physical volumes, but other OSs may not recognize

the LVM metadata, and display the drive as free and possibly overwrite it.

4.3 Creating Volume Groups

To create a volume group, the following command is used:

# vgcreate <vg_name > <device>

4.4 Creating Logical Volumes

To create a logical volume within a volume group, the following command can be used:

# lvcreate -L <size> --name <lv_name> <vg_name>

4.5 Deleting Logical Volumes

Warning: Deleting a logical volume cannot be reverted. All data on this logical volume will

be lost. Please make sure to unmount the logical volume before deleting it.

# umount <mount_point>

# lvremove -f <vg_name>/<lv_name>

4.6 Resizing Logical Volumes

If a logical volume needs to be resized, it is recommended to logout any attached initiators from this

device, as an initiator relogin is necessary to provide the new block device specifications.

# lvresize -L+1G <vg_name>/<lv_name>

剩余81页未读，继续阅读

懂你668

粉丝: 1
资源: 8

Datera LIO Linux SCSI Target 管理手册

LIO-SAM:LIO-SAM

对于fast-lio论文和代码的一些个人理解

FAST_LIO_SAM-fast-lio

轮廓 LIO User Guide for Contour LIO Data Class

Lio4客户端

zigzag代码matlab-LIO-SAM:LIO-SAM

targetcli-fb:用于管理Linux LIO内核目标的命令外壳

LIO:查找对象

fast lio和lio sam算法优劣性对比

fast lio和lio sam算法的优劣对比

最新资源