20160127-VMware-VSOS-Day 3
Lesson 3: Troubleshooting Network Performance Problems
Dropped Receive Packets
- Cause
- High CPU utilization
- Improper guest operation system driver configuration
Dropped Transmit Packets
- Cause
- Traffic exceeds the physical capabilities of the uplink NICs or the networking infrastructure.
Module 7 Storage Scalability
Lesson 1: Storage APIs and Virtual Machine Storage Policies
VMFS-3
- Block size / Volume size
- 1MB / 256GB
- 2MB / 512GB
- 4MB / 1TB
- 8MB / 2TB
- 單檔 2TB
VMFS-5
- Block size / Volume size
- 1MB / 64TB
- ESXi 5.0 / 5.1 單檔 2TB
- ESXi 5.5 / 6.0 單檔 62TB
SCSI Reservations
- 當ESXi 作5項工作時, 會短暫 Lock file system
- Creating or deleting a virtual disk
- Increasing the size of a VMFS volume
- Creating or deleting a snapshots
- Increasing the size of a VMDK file
- 支援VAAI 的儲存設備不受此限制
Virtual Disk Types
- Eager-zeroed thick
- 建立時直接佔滿所需空間
- Quorum drive in an MSCS cluster, Fault Tolerance
- Lazy-zeroed thick
- 建立時直接佔滿所需空間
- Good for most cases.
- Thin
- 建立時檔案很小, 寫入資料後會變大
- Disk space utilization is the main concern.
About vSphere Storage APIs - Array Integration
VAAI 具備
- 硬體加速
- 在storage 內進行, 不經由 ESXi
- Thin Provisioning API ( 實際空間小於宣告的空間 )
- 偵測使用空間, 提醒管理者要增加實際空間
- 實際上應該不用如此, 如果 LUN 空間變大之後, 只要到 datastore 上面設定擴充
About vSphere API for Storage Awareness (VASA)
- Storage Provide
- 確認storage 是否有支援 VAAI 或是相關功能
- 在 storage -- > Manage -- > Storage Providers
-- Class break --
About Virtual Machine Storage Policies
Lab 6: Policy-Based Storage
- 參考 Lab 文件
- Task 1: Add Datastores for Use by Policy-Based Storage
- Task 2: Use vSphere vMotion to Migrate a Virtual Machine to the Gold Datastore
- Task 3: Configure Storage Tags
- Task 4: Create Virtual Machine Storage Policies
- Task 5: Assign Storage Policies to Virtual Machines
Notes
- Storage Policy 要搭配 VASA Storage Provider 的功能, 才能設定裝置上面的 Read 或是 Write 相關條件, 不然只是 Tag 功能, 實際意義不大
Lesson 2: vSphere Storage I/O Control
Automatic Threshold Detection
- 預設 30 ms
- 或是 Throughput 90% 的值
vSphere Storage I/O Control Requirements
- Support Fibre Channel, iSCSI, and NFS
- Raw device (RDM) is not supported
- Does not support datastores with multiple extents. ( VMFS 跨多個 LUN )
Lesson 3: Datastore Clusters and vSphere Storage DRS
Datastore Cluster
- 不是容錯系統
- 一群 datastore 的定義
- 在此範圍內做 Storage DRS
Datastore Cluster Requirements
- 用類似的datastore 組成
- ESXi 必須 5.0 以上
- 不可以跨 data centers ( 考量速度? )
Storage Migration Recommendations
- Executed under the following circumstances
- When the I/O latency threshold is exceeded
- IOPS load history is checked every 8 hours by default
- When the space utilization threshold is exceeded
- checked every 5 mins by default
vSphere Storage DRS Maintenance Mode
- 當 Storage Cluster 內的某個 datastore 進 maintenance mode
- Registered virtual machines ( on or off ) are moved
- Templates, unregistered virtual machines, ISO images, and non-virtual machine files are not moved.
-- 中午休息 --
Lab 7: Managing Datastore Clusters
- 參考 Lab 文件
- Task 1: Create a Datastore Cluster That is enable by vSphere Storage DRS
- Task 2: Evacuate a Datastore Using Datastore Maintenance Mode
- Task 3: Run vSphere Storage DRS and Apply Migration Recommendations
- Task 4: Clean Up for Next Lab
Flash Read Cache ( Web Client 上面顯示 Virtual Flash )
- 插入SSD
- 參考 Slide 7-59 在主機 Manage -- > Setting -- > Virtual Flash, 設定 Flash Read Cache 容量
- 於 VM -- > Edit Settings -- > Hard Disk
Flash Read Cache Limitations
- Support for only locally attached SSDs
- Not compatible with VMware FT
- Cannot share an SSD with Virtual SAN or a VMFS datastore.
Virtual SAN
- ESXi 內建功能
- 另購 License
- 3 ~ 64 Hosts / Cluster
- 只能存 VM 的檔案
- VM 必須套用 VM Storage Policy
Virtual Volumes
- 在 vCenter 上直接管 storage, 需要有 storage provider
-- Class break --
Lab 8: Working with Virtual Volumes
- 參考 Lab 文件
- Task 1: Validate the Storage Endpoint
- Task 2: Create a vSphere Virtual Volumes Datastore Using NAS
- Task 3: Create a vSphere Virtual Volumes Datastore Using iSCSI
- Task 4: Migrate a Virtual Machine to a Virtual Volume
Module 8 Storage Optimization
之後再提到
Module 9 CPU Optimization
LCPU 數量 = Socket 數量 * core數 * HT(2)
CPU Performance Factors
- 如果給 VM 過多的 vCPU
- 浪費
- 閒置或是低使用量, 會造成一直做 CPU Load Balancing
- License
Ready Time
- vCPU 等候可用的LCPU所需的時間, 判斷實體 CPU 數量比率是否足夠
Lesson 2: Monitoring CPU Activity
Notes
- Ready Time 大約等於 Used time 表示 Okay
Lab 10: Monitoring CPU Performance
- 參考 Lab 文件, 先做到 Task 6
- Task 1: Run a Single-Threaded Program in a Single-vCPU Virtual Machine
- Task 2: Start resxtop and Record Baseline Statistics
- Task 3: Configure an Advance Performance Chart and Record Baseline Statistics
- Task 4: Import the CPU-HOG vAPP
- Task 5: Generate CPU Contention
- Task 6: Record Statistics for Case 1: Single Thread and Single vCPU
沒有留言:
張貼留言