guide line
best practise
新建cluster注意事项:
- 指定独立的VNet,保证可以做灵活的网络管控
- 对于节点多的cluster,为了保证Ambari监控性能,最好申请独立的DB,并且提高其等级,机器越多建议提高的等级越多。
azure cosmos命令
d:\project\APGold\autopilotservice\Global\VirtualEnvironments\AdsBI\AdsMz-Test-MW1-MSNMediation>sd edit deployment.int
deployment.int - file(s) not on client.
直接编辑
codeflow作用
ve作用:
pe作用:
之前
D:\project\APGold\autopilotservice\Global\VirtualEnvironments\AdsBI\AdsMz-Test-MW1-MSNMediation
D:\project\APGold\autopilotservice\MW1\AdsMz-Prod-MW1
A ServiceGroup is mandatory for every machine function
VE中存储的可以理解为某个项目的global config,所有的PE公用这部分config
设计目的是在同一项目部署多个PE时,避免多次发布,只需要更新VE,相关PE就会生效
https://msazure.visualstudio.com
D:\project\APGold\autopilotservice\Global\VirtualEnvironments\AdsBI\AdsMz-Prod-MW1-MSNMediation
1 | 07/26/21 16:54:01.841,error processing AdsMz-Test-MW1: Rollout 'AdsBI-AdsMz-Test-MW1-MzOrchestration-VE.6799404_17553440741849145769_0.Rollout_AdsBI_Streaming_CFR' cannot be kicked due to status: InRollout - Rollout in progress..Rollout 'AdsBI-AdsMz-Test-MW1-MSNMediation-VE.6827694_4064931696698155710_0.Rollout_AdsBI_Streaming_CFR' cannot be kicked due to status: OtherRollout - Waiting on other rollout on MF AdsBI_Streaming_CFR in progress: AdsBI_Streaming_CFR.AdvertiserAggs_94d0ef91_90546_1_15-0-897+94d0ef91_CL6670088_5920165827906938206_0.AdsBI-AdsMz-Test-MW1-AdvertiserAggs-VE.csv:AdsBI_Streaming_CFR.AgoraAggs_76700988_85934_1_16-0-173+76700988_CL6670088_4056715873563314682_0.AdsBI-AdsMz-Test-MW1-AgoraAggs-VE.csv:AdsBI_Streaming_CFR.CACFR_9a63868e_90143_2_1-0-2408+9a63868e_CL6670088_11167218336511992544_0.AdsBI-AdsMz-Test-MW1-CACFR-VE.csv:AdsBI_Streaming_CFR.CFR_2d2cea68_91067_1_17-0-6662+2d2cea68_CL6670088_18415503406869780426_0.AdsBI-AdsMz-Test-MW1-CFR-VE.csv:AdsBI_Streaming_CFR.KpiAggs_fdd5f92d_90180_1_17-0-433+fdd5f92d_CL6670088_6626197140932378707_0.AdsBI-AdsMz-Test-MW1-KPIAggs-VE.csv:AdsBI_Streaming_CFR.Monetization_de0db608_91071_1_1-0-2808+de0db608_CL6670088_351842135959690630_0.AdsBI-AdsMz-Test-MW1-Monetization-VE.csv:AdsBI_Streaming_CFR.Orchestration_93eb645e_91078_1_16-0-557+93eb645e_CL6799404_17553440741849145769_0.AdsBI-AdsMz-Test-MW1-MzOrchestration-VE.csv:AdsBI_Streaming_CFR.PublisherAggs_55bd6dd9_90534_1_17-0-474+55bd6dd9_CL6670088_15361642287162420254_0.AdsBI-AdsMz-Test-MW1-PublisherAggs-VE.csv:AdsBI_Streaming_CFR.Mediation_865d3b39_90170_1_merge_20210629_-1_CL6670088_3189803787641037971_0.AdsBI-AdsMz-Test-MW1-Mediation-VE.csv:AdsBI_Streaming_CFR.MSNMediation_4a63e32e_91077_2_0-1-28+4a63e32e33_CL6827694_4798040603905411799_0.AdsBI-AdsMz-Test-MW1-MSNMediation-VE.csv:.OM.Autopilot-AutopilotClient-VE.csv..; .. Some rollouts not triggered. |
1 | Deployment: 6827694_4798040603905411799 (this is NOT the latest deployment.ini change number) |
1 | Deployment: 6827694_9979618793807238839_0 (this is NOT the latest deployment.ini change number) |
Severity Code Description Project File Line Suppression State
Error MSB3680 The source file “D:\project\Ads.BI.MSNMediation\private\src\Batch/Autopilot/ExternalConfigs/Microsoft.BingAds.BI.APDeploy.exe.config” does not exist. Microsoft.BI.MSNMediation.HourlyS2B.Drop D:\project\Ads.BI.MSNMediation\private\src\Batch\build\targets\CreateDrop.targets 47
、
Severity Code Description Project File Line Suppression State
Error MSB3030 Could not copy the file “D:\project\Ads.BI.MSNMediation\private\src\Batch/SAMHourlyS2B/Microsoft.BI.MSNMediation.HourlyS2B.Drop/bin\Debug\net472**” because it was not found. Microsoft.BI.MSNMediation.HourlyS2B.Drop D:\project\Ads.BI.MSNMediation\private\src\Batch\build\targets\CreateDrop.targets 47
c:\Users\jingqicao.nuget\packages\microsoft.bingads.bi.apdeploy\10.5.2
在sdk之外做的nuget package改动,比如删掉某个包
resolved Show files that have been merged but not submitted 跟执行目录无关
retype Reappraise the file type for files on the client
revert Discard changes from an opened file
review List and track changelists (for the review daemon)
sd review 如何只搜索某个人的review
FAREAST\jingqicao
sd opened
sd submit -c
scope插件安装
https://mscosmos.visualstudio.com/CosmosWiki/_wiki/wikis/Cosmos.wiki/66/SCOPE-Tutorial
时区
服务器时间: UTC
xts log中打印的时间:UTC
xts log构建时间(xts服务器时间):美国时间
puid
gid 999+工号
company simple: MSFT
ESPP
R开头的一串数字
HRWeb
信息维护,会影响摩根账户,需要全部都是英文,包括地址
https://myaccount.microsoft.com/?ref=MeControl
中意人寿
family
143cc3b8f01f
https://apkpure.com
步骤:Onboard AP App Deployment using AzDeployer (Stratus)
1 | ##[warning]C:\Program Files (x86)\Microsoft Visual Studio\2019\Enterprise\MSBuild\Current\Bin\amd64\Microsoft.Common.CurrentVersion.targets(2203,5): Warning MSB3245: Could not resolve this reference. Could not locate the assembly "Google.Protobuf, Version=3.6.1.0, Culture=neutral, PublicKeyToken=a7d26565bac4d604, processorArchitecture=MSIL". Check to make sure the assembly exists on disk. If this reference is required by your code, you may get compilation errors. |
1 | 2021-08-02 06:33:05 ThreadId: 69 PauseAndResumeWorkflows() Process: MSNBI_S2BHourly_0 UserCommand: Run |
1 | 2021-08-02T08:28:02.5629452Z Checking compatibility for System.Security.Cryptography.Primitives 4.3.0 with .NETFramework,Version=v4.7.2. |
ads Data live-site tracking v2
Deployment process for FASTBI streaming pipeline - V2
[toc]
这是一个很专业的问题。你观察到的这种“快速定位并总结参考文献”的能力,并非单个模型的魔法,而是一套精密系统协同工作的结果。这套系统通常被称为 RAG(检索增强生成)。
简单来说,整个过程可以分解为三个核心阶段,每个阶段都有专门的功能模块参与。
这个阶段的目标是把你的口语化问题,变成系统能精准检索的“钥匙”。
这是实现“快速”的关键。系统不会去一篇篇阅读原始文档,而是通过“语义指纹”来匹配。
系统不是简单地把文章拼起来,而是让大模型扮演一个“擅长总结的编辑”。
核心模块:大语言模型
构建上下文:系统会把你原始的问题,和阶段二找到的最相关的几篇文章(或段落),一起打包放进一个精心设计的提示模板里。比如:
“根据以下参考资料,用中文简洁、准确地回答用户的问题。如果资料里没有答案,就说不知道。\n参考资料:\n[文章1内容]\n[文章2内容]\n…\n用户问题:[你的问题]”
生成回答:大模型看到这个提示后,会“根据”提供的资料来生成答案。它会提取关键信息、理顺逻辑、用自己的话进行总结,并在结尾附上参考来源。
流式输出:生成答案的同时,系统以“打字机效果”逐字返回,让你感觉响应极快。
假设你问:“2024年诺贝尔物理学奖得主是谁?”
这套RAG架构,就像给大模型外挂了一个可以无限扩展的、能极速检索的“图书馆”,让它既能博闻强识(访问最新外部知识),又能避免胡编乱造(回答基于事实资料)。这已经成为目前解决大模型知识更新和幻觉问题的主流工业方案。