SLES 12 for SAP 上課小記
Day 1:
下載上課檔案
安裝 7-zip
# zypper install p7zip
解壓縮檔案
# 7z x SLE321.7z.001
進入相關目錄
# cd SLE321/
調整 install script 的權限
# chmod a+x install_lab_env.sh
執行環境檢查
# ./ install_lab_env.sh
如果不符合環境的要求, 會出現相關修改建議
例如我這邊就是要修改 /etc/libvirt/libvirtd.conf 相關設定
# vi /etc/libvirt/libvirtd.conf
Locate and uncomment/edit the following line to match:
unix_sock_group = libvirt
unix_sock_ro_perms = "0777"
unix_sock_rw_perms = "0770"
unix_sock_dir = "/var/run/libvirt"
auth_unix_ro = "none"
auth_unix_rw = "none"
listen_tcp = 1
Locate and uncomment/edit the following line to match:
auth_tcp = "none"
重新開機
# reboot
進入相關目錄
# cd SLE321/
執行環境檢查
# ./ install_lab_env.sh
新增檔案並加入下列設定
# vi /etc/profile.d/libvirt.sh
Add the following line to the end of the file:
export LIBVIRT_DEFAULT_URI="qemu:///system"
登出環境
# cd SLE321/
執行環境檢查
# ./ install_lab_env.sh
將 root 加入 libvirt 群組
# usermod -aG libvirt root
登出環境
# cd SLE321/
執行環境檢查
# ./ install_lab_env.sh
完成環境安裝
Cluster Terminology
- Epoch - The version number of the cluster configuration / status
- Split Brain
- A state where the nodes in a cluster disagree about the current state of the cluster. Must be resolved before the cluster can continue to operate
- Fencing
- Method to protect the cluster by denying access to resources or rebooting a node
- 隔離, 用來防止 Split Brain
- Active / Passive
- Active / Active
The SLE High Availability Extension
- Beginning with SLES 11, the cluster packages are no longer included in the base product
- Instead, the cluster stack is available as an add-on product, or extension
- SLES 12 integrates the delivery of the HA Extension into the software manager
- A separate subscription is required for each node that will run the HA Extension
- Up to 32 nodes
- Geo Cluster Extension to add multisite support
列出 pattern
# zypper pt
Deployment Options
- Cluster Types
- Local Cluster
- Metro Cluster
- Geo Cluster
Main components of HAE
- High availability base components
- Corosync - 心跳, 透過 totem 實現同步
- Pacemaker - Cluster 的大腦
- Management tool
- crm shell ( CRM - Cluster Resource Manager )
- HAWK / HAWK2
- Stonith
- 用來進行隔離(Fencing)
- Shoot The Other Node In the Head
- CIB
- Cluster Information Base
- 用來存放目前 Cluster 目前運作的狀態
- 只有主節點會更新.
- PE
- Policy Engine, 用來計算事件如何應對
- 運作在 DC 節點
- LRM
- Local Resource Manager
- 執行 CRM 發出的指令
- RA
- Resource Agent
- DLM
- Distributed Lock Manager
- Provide generic cluster wide locking facilities
- Storage components
- cLVMd
- 在 cluster 用 LVM
- DRBD
- (Distributed Replicated Block Device) 是 Linux 平台上的分散式儲存系統
- 不提供同時讀寫
- OCFS2 filesystem
- GFS filesystem
- Load balancing
- Linux IP Virtual Server ( IPVS )
- HAProxy
- Conntrackd
如何判斷主節點?
- Cluster 內其實沒有主節點或是副節點,
- 有DC 節點也許可以視為主節點, 上面的 PE
==== Break ====
Cluster Resource Manager ( CRM )
讓大家完成環境建置
==== Lunch Break ====
Lab1: Power On and Verify the Storage Server
Lab2: Install and Configure a Cluster
Section 3: Cluster Administration Tools
HAWK
Lab 3: Cluster Administrator Tools
透過 HAWK2 以 web去觀察 cluster
透過 crm 指令去觀察與控制 cluster
透過 pssh 指令去平行處理 ssh 指令
透過 csync2 指令去同步各個 node 的檔案