【Survey】 NVIDIA Clara Deploy SDK - 安裝篇

Deploy SDK Part2！

原本想在上一篇一起解決的，不過最後範例程式碼架上選染效果後真的超級長的，最後決定來開一篇新的寫安裝步驟好了。

但是說這篇真的拖有夠久的，產出速度比不上待寫文章的新增速度，結果草稿越積越多 Orz

安裝篇 （圖片來源: Meet創業小聚）

按照文件看來安裝步驟似乎不難，不過按照以往經驗看來成不成功要看人品與緣份 XDDD

系統需求

在文件一開始列了落落長的系統需求，除了一些基本硬體需求外，還有指定了 K8S、 Helm 與 Docker 的版號：

• Kubernetes 1.15.4
• Docker 19.03.1
• NVIDIA Docker 2.2.0
• Helm 2.15.2

如果系統中未安裝這些需求，可以不用先行安裝，等等安裝時會幫忙一併安裝；但如果安裝的版號不合，可能需要先行移除，否則它們會跳過安裝。

安裝步驟

上張圖說明一下 Deploy SDK 的安裝流程：

安裝流程 （圖片來源: SDK 0.7.1 documentation）

1. 下載並安裝 bootstrap
   首先先登入 NGC，找到 Clara Deploy Bootstrap 並進行下載與解壓縮：

   1 2	$ wget --content-disposition https://api.ngc.nvidia.com/v2/resources/nvidia/clara/clara_bootstrap/versions/0.7.1-2008.1/zip -O bootstrap.zip $ unzip bootstrap.zip -d bootstrap

   完成下載後，進入資料夾執行腳本。這份腳本它將會安裝 Docker、 K8S …等所需求的軟體：

   1 2	$ cd bootstrap $ sudo ./bootstrap.sh

   是說，如果不想登入 NGC 也行，登入與否其實不影響下載。不過還是建議登入下，否則很有機會在接下來的步驟中被它打斷，它實在很吵…

   
   

2. 下載並安裝 CLI
   接下來再去 NGC，找 Clara CLI 來下載與解壓縮：

   1 2 3 4 5 6 7 8 9

   $ wget --content-disposition https://api.ngc.nvidia.com/v2/resources/nvidia/clara/clara_cli/versions/0.7.1-2008.1/zip -O clara_cli.zip $ sudo unzip clara_cli.zip -d /usr/bin/ && sudo chmod 755 /usr/bin/clara* Archive: cli.zip inflating: /usr/bin/clara inflating: /usr/bin/clara-dicom inflating: /usr/bin/clara-monitor inflating: /usr/bin/clara-pull inflating: /usr/bin/clara-render

   將檔案放到 /usr/bin/ 下後，可以試著呼叫 clara 指令，驗證是否安裝成功：

   1 2	$ clara version Clara CLI version: 0.7.1-12788.ae65aea0

   
   

3. 配置 NGC 憑證
   安裝 Clara CLI 須配置 NGC 憑證，稍等 Clara CLI 才能從 NGC Pull 相關 Helm Chart 以進行部署。

   這邊你須要拿到一把 NGC_API_KEY。這次就必須一定要登入 NGC 了，登入後點選右上角頭像選單中的 Setup，並選擇 Generate API Key

   

   進入頁面後，會右上方有個 Generate API Key 的按鈕，點擊後就會產生 NGC_API_KEY 了。

   

   完成後回到終端輸入下列指令，可以考慮 orgteam 使用預設值就好，：

   1 2 3

   $ clara config --key NGC_API_KEY [--orgteam nvidia/clara] -y ✔ Yes Configuration "ngc-clara"successfully created

   
   

   是說 successfully 的意思，是指你成功配置了憑證，但憑證是否能使用必須使用後才知道，可以試著使用 pull 指令來試試：

   1 2 3 4 5

   $ clara pull platform ✔ Yes Clara Platform 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara Hint: use "clara platform start" or "clara platform restart" to deploy pulled Clara Platform.

   如果失敗可能會看到下列這樣的訊息：

   1 2 3 4

   $ clara pull platform ✔ Yes Error: unable to fetch latest version information 401 Unauthorized

   或是

   1 2 3 4

   $ clara pull platform ✔ Yes Error: unable to fetch latest version information 403 Forbidden

   
   

4. 啟動 Helm Chart
   在上一篇提到 Helm Charts 時有提過，除了 Triton Inference Server 之外的 charts，都可以藉由這步驟啟動。

   DNVIDIA Clara Deploy Architecture（圖片來源: SDK 0.7.1 documentation）

   因為 platform 的下載在剛剛測試 clara 的指令時已經順道完成了，所以這邊就直接啟動。

   1 2 3 4

   $clara platform start Starting clara... NAME: clara Note: If there is a running instance of Clara Console, Clara Dicom Adapter or Clara Renderer, they should be restarted.

   
   

   接下來下載 Clara Deploy Services 的 Helm Charts：

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

   $ clara pull dicom ✔ Yes Clara Dicom Adapter 0.7.1-2008.1 Chart saved at: /home/.clara/charts/dicom-adapter Hint: use "clara dicom start" or "clara dicom restart" to deploy pulled Clara Dicom Adapter. $ clara pull render ✔ Yes Clara Renderer 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara-renderer Hint: use "clara render start" or "clara render restart" to deploy pulled Clara Renderer. $ clara pull monitor ✔ Yes Clara Monitor Server 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara-monitor-server Hint: use "clara monitor start" or "clara monitor restart" to deploy pulled Clara Monitor Server. $ clara pull console ✔ Yes Clara Management Console 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara-console Hint: use "clara console start" or "clara console restart" to deploy pulled Clara Management Console.

   之後就可以試著啟動了：

   1 2 3 4 5 6 7 8 9 10 11 12 13

   $ clara dicom start Starting DICOM Adapter... NAME: clara-dicom-adapter $ clara render start NAME: clara-render-server $ clara monitor start NAME: clara-monitor-server $ clara console start NAME: clara-console

驗證安裝

如果一切順利的話，跑完上面算是安裝完成了，你可以試著下 hlem ls 指令來觀察目前所啟動的 charts：

1 2 3 4 5 6 7

$ helm ls NAME CHART clara clara-0.7.1-2008.1 clara-console clara-console-0.7.1-2008.1 clara-dicom-adapter dicom-adapter-0.7.1-2008.1 clara-monitor-server clara-monitor-server-0.7.1-2008.1 clara-render-server clara-renderer-0.7.1-2008.1

或是下 kubectl get pods 應該會看到下面這些 Pods：

• clara-clara-platformapiserver-
• clara-dicom-adapter-
• clara-monitor-server-fluentd-elasticsearch-
• clara-monitor-server-grafana-
• clara-monitor-server-monitor-server-
• clara-render-server-clara-renderer-
• clara-resultsservice-
• clara-ui-
• clara-console-
• clara-console-mongodb-
• clara-workflow-controller-
• elasticsearch-master-0
• elasticsearch-master-1

觀察 Pod 的變化

提到啟動的 Pod，有點好奇在每個 Chart 啟動時，會啟動的 Pod 有哪些。所以把整個 Clara 卸掉，重新安裝一次並觀察 Pod 的變化。

1. 安裝前

   1 2 3 4 5 6

   $ kubectl get all NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 4d23h $ kubectl get pods No resources found.

   
   

2. clara platform start

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30

   $ kubectl get all NAME READY STATUS RESTARTS AGE pod/clara-clara-platformapiserver-54c5c44bbd-9b97b 1/1 Running 0 95s pod/clara-resultsservice-664477898f-zl8cr 1/1 Running 0 95s pod/clara-ui-6f89b97df8-fn2zm 1/1 Running 0 95s pod/clara-workflow-controller-69cbb55fc8-t67ns 1/1 Running 0 95s pod/fluentd-7n2b8 1/1 Running 0 95s pod/fluentd-ccnzw 1/1 Running 0 95s NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE service/clara NodePort 10.103.37.52 <none> 50051:31536/TCP 95s service/clara-resultsservice ClusterIP 10.108.91.220 <none> 8088/TCP 95s service/clara-ui ClusterIP 10.97.148.11 <none> 80/TCP 95s service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 9m52s NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE daemonset.apps/fluentd 2 2 2 2 2 <none> 95s NAME READY UP-TO-DATE AVAILABLE AGE deployment.apps/clara-clara-platformapiserver 1/1 1 1 95s deployment.apps/clara-resultsservice 1/1 1 1 95s deployment.apps/clara-ui 1/1 1 1 95s deployment.apps/clara-workflow-controller 1/1 1 1 95s NAME DESIRED CURRENT READY AGE replicaset.apps/clara-clara-platformapiserver-54c5c44bbd 1 1 1 95s replicaset.apps/clara-resultsservice-664477898f 1 1 1 95s replicaset.apps/clara-ui-6f89b97df8 1 1 1 95s replicaset.apps/clara-workflow-controller-69cbb55fc8 1 1 1 95s

   
   

3. clara dicom start

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40

   $ kubectl get all NAME READY STATUS RESTARTS AGE pod/clara-clara-platformapiserver-54c5c44bbd-9b97b 1/1 Running 0 2m44s pod/clara-dicom-adapter-7948fcd445-rtbqr 1/1 Running 0 33s pod/clara-resultsservice-664477898f-zl8cr 1/1 Running 0 2m44s pod/clara-ui-6f89b97df8-fn2zm 1/1 Running 0 2m44s pod/clara-workflow-controller-69cbb55fc8-t67ns 1/1 Running 0 2m44s pod/fluentd-7n2b8 1/1 Running 0 2m44s pod/fluentd-ccnzw 1/1 Running 0 2m44s NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) $ GE service/clara NodePort 10.103.37.52 <none> 50051:31536/TCP $ m44s service/clara-dicom-adapter NodePort 10.105.101.54 <none> 104:31289/TCP,5000:31880/TCP $ 3s service/clara-resultsservice ClusterIP 10.108.91.220 <none> 8088/TCP $ m44s service/clara-ui ClusterIP 10.97.148.11 <none> 80/TCP $ m44s service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP $ 1m NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE daemonset.apps/fluentd 2 2 2 2 2 <none> 2m44s NAME READY UP-TO-DATE AVAILABLE AGE deployment.apps/clara-clara-platformapiserver 1/1 1 1 2m44s deployment.apps/clara-dicom-adapter 1/1 1 1 33s deployment.apps/clara-resultsservice 1/1 1 1 2m44s deployment.apps/clara-ui 1/1 1 1 2m44s deployment.apps/clara-workflow-controller 1/1 1 1 2m44s NAME DESIRED CURRENT READY AGE replicaset.apps/clara-clara-platformapiserver-54c5c44bbd 1 1 1 2m44s replicaset.apps/clara-dicom-adapter-7948fcd445 1 1 1 33s replicaset.apps/clara-resultsservice-664477898f 1 1 1 2m44s replicaset.apps/clara-ui-6f89b97df8 1 1 1 2m44s replicaset.apps/clara-workflow-controller-69cbb55fc8 1 1 1 2m44s

   
   

4. clara render start

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46

   $ kubectl get all kubectl get all NAME READY STATUS RESTARTS AGE pod/clara-clara-platformapiserver-54c5c44bbd-gfwng 1/1 Running 0 24m pod/clara-dicom-adapter-7948fcd445-mv248 1/1 Running 0 20m pod/clara-render-server-clara-renderer-d79dd4779-f5hgd 2/3 CrashLoopBackOff 7 11m pod/clara-resultsservice-664477898f-2vsw9 1/1 Running 0 24m pod/clara-ui-6f89b97df8-c5p2f 1/1 Running 0 24m pod/clara-workflow-controller-69cbb55fc8-mc682 1/1 Running 0 24m pod/fluentd-ntl6q 1/1 Running 0 24m pod/fluentd-tvnrl 1/1 Running 0 24m NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE service/clara NodePort 10.101.135.71 <none> 50051:32455/TCP 24m service/clara-dicom-adapter NodePort 10.100.25.126 <none> 104:31985/TCP,500 0:30647/TCP 20m service/clara-renderer-clara-render-server NodePort 10.108.60.232 <none> 8070:30105/TCP,80 60:32006/TCP 11m service/clara-resultsservice ClusterIP 10.109.206.204 <none> 8088/TCP 24m service/clara-ui ClusterIP 10.101.195.91 <none> 80/TCP 24m service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 27m NAME DESIRED CURRENT READY UP-TO-DATE AVAILABLE NODE SELECTOR AGE daemonset.apps/fluentd 2 2 2 2 2 <none> 24m NAME READY UP-TO-DATE AVAILABLE AGE deployment.apps/clara-clara-platformapiserver 1/1 1 1 24m deployment.apps/clara-dicom-adapter 1/1 1 1 20m deployment.apps/clara-render-server-clara-renderer 0/1 1 0 11m deployment.apps/clara-resultsservice 1/1 1 1 24m deployment.apps/clara-ui 1/1 1 1 24m deployment.apps/clara-workflow-controller 1/1 1 1 24m NAME DESIRED CURRENT READY AGE replicaset.apps/clara-clara-platformapiserver-54c5c44bbd 1 1 1 24m replicaset.apps/clara-dicom-adapter-7948fcd445 1 1 1 20m replicaset.apps/clara-render-server-clara-renderer-d79dd4779 1 1 0 11m replicaset.apps/clara-resultsservice-664477898f 1 1 1 24m replicaset.apps/clara-ui-6f89b97df8 1 1 1 24m replicaset.apps/clara-workflow-controller-69cbb55fc8 1 1 1 24m

   
   

5. clara monitor start

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70

   $ kubectl get all NAME READY STATUS RESTARTS AGE pod/clara-clara-platformapiserver-54c5c44bbd-gfwng 1/1 Running 0 40m pod/clara-dicom-adapter-7948fcd445-mv248 1/1 Running 0 36m pod/clara-monitor-server-fluentd-elasticsearch-dl7bj 1/1 Running 0 14m pod/clara-monitor-server-fluentd-elasticsearch-jxdk6 1/1 Running 0 14m pod/clara-monitor-server-grafana-5f874b974d-qvxgn 1/1 Running 0 14m pod/clara-monitor-server-monitor-server-59c8bf68f7-5rcg7 0/1 CrashLoopBackOff 7 14m pod/clara-render-server-clara-renderer-d79dd4779-f5hgd 2/3 CrashLoopBackOff 10 27m pod/clara-resultsservice-664477898f-2vsw9 1/1 Running 0 40m pod/clara-ui-6f89b97df8-c5p2f 1/1 Running 0 40m pod/clara-workflow-controller-69cbb55fc8-mc682 1/1 Running 0 40m pod/elasticsearch-master-0 1/1 Running 0 14m pod/elasticsearch-master-1 1/1 Running 0 14m pod/fluentd-ntl6q 1/1 Running 0 40m pod/fluentd-tvnrl 1/1 Running 0 40m NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE service/clara NodePort 10.101.135.71 <none> 50051:32455/TCP 40m service/clara-dicom-adapter NodePort 10.100.25.126 <none> 104:31985/TCP,500 0:30647/TCP 36m service/clara-monitor-server ClusterIP 10.111.167.160 <none> 50051/TCP 14m service/clara-monitor-server-grafana NodePort 10.100.148.116 <none> 80:32000[16/1632] 14m service/clara-renderer-clara-render-server NodePort 10.108.60.232 <none> 8070:30105/TCP,80 60:32006/TCP 27m service/clara-resultsservice ClusterIP 10.109.206.204 <none> 8088/TCP 40m service/clara-ui ClusterIP 10.101.195.91 <none> 80/TCP 40m service/elasticsearch-master ClusterIP 10.108.240.18 <none> 9200/TCP,9300/TCP 14m service/elasticsearch-master-headless ClusterIP None <none> 9200/TCP,9300/TCP 14m service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 43m NAME DESIRED CURRENT READY UP-TO-DATE AVAI LABLE NODE SELECTOR AGE daemonset.apps/clara-monitor-server-fluentd-elasticsearch 2 2 2 2 2 <none> 14m daemonset.apps/fluentd 2 2 2 2 2 <none> 40m NAME READY UP-TO-DATE AVAILABLE AGE deployment.apps/clara-clara-platformapiserver 1/1 1 1 40m deployment.apps/clara-dicom-adapter 1/1 1 1 36m deployment.apps/clara-monitor-server-grafana 1/1 1 1 14m deployment.apps/clara-monitor-server-monitor-server 0/1 1 0 14m deployment.apps/clara-render-server-clara-renderer 0/1 1 0 27m deployment.apps/clara-resultsservice 1/1 1 1 40m deployment.apps/clara-ui 1/1 1 1 40m deployment.apps/clara-workflow-controller 1/1 1 1 40m NAME DESIRED CURRENT READY AGE replicaset.apps/clara-clara-platformapiserver-54c5c44bbd 1 1 1 40m replicaset.apps/clara-dicom-adapter-7948fcd445 1 1 1 36m replicaset.apps/clara-monitor-server-grafana-5f874b974d 1 1 1 14m replicaset.apps/clara-monitor-server-monitor-server-59c8bf68f7 1 1 0 14m replicaset.apps/clara-render-server-clara-renderer-d79dd4779 1 1 0 27m replicaset.apps/clara-resultsservice-664477898f 1 1 1 40m replicaset.apps/clara-ui-6f89b97df8 1 1 1 40m replicaset.apps/clara-workflow-controller-69cbb55fc8 1 1 1 40m NAME READY AGE statefulset.apps/elasticsearch-master 2/2 14m

   
   

6. clara console start

   1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81

   $ kubectl get all NAME READY STATUS RESTARTS AGE pod/clara-clara-platformapiserver-54c5c44bbd-gfwng 1/1 Running 0 61m pod/clara-console-8565b4d565-77jhc 2/2 Running 0 19m pod/clara-console-mongodb-85f8bd5f95-8nwqx 1/1 Running 0 19m pod/clara-dicom-adapter-7948fcd445-mv248 1/1 Running 0 58m pod/clara-monitor-server-fluentd-elasticsearch-dl7bj 1/1 Running 0 36m pod/clara-monitor-server-fluentd-elasticsearch-jxdk6 1/1 Running 0 36m pod/clara-monitor-server-grafana-5f874b974d-qvxgn 1/1 Running 0 36m pod/clara-monitor-server-monitor-server-59c8bf68f7-5rcg7 0/1 CrashLoopBackOff 11 36m pod/clara-render-server-clara-renderer-d79dd4779-f5hgd 2/3 CrashLoopBackOff 14 48m pod/clara-resultsservice-664477898f-2vsw9 1/1 Running 0 61m pod/clara-ui-6f89b97df8-c5p2f 1/1 Running 0 61m pod/clara-workflow-controller-69cbb55fc8-mc682 1/1 Running 0 61m pod/elasticsearch-master-0 1/1 Running 0 36m pod/elasticsearch-master-1 1/1 Running 0 36m pod/fluentd-ntl6q 1/1 Running 0 61m pod/fluentd-tvnrl 1/1 Running 0 61m NAME TYPE CLUSTER-IP EXTERNAL-IP PORT(S) AGE service/clara NodePort 10.101.135.71 <none> 50051:32455/TCP 61m service/clara-console NodePort 10.99.119.217 <none> 8080:32002/TCP,50 00:32003/TCP 19m service/clara-console-mongodb ClusterIP 10.102.177.195 <none> 27017/TCP 19m service/clara-dicom-adapter NodePort 10.100.25.126 <none> 104:31985/TCP,500 0:30647/TCP 58m service/clara-monitor-server ClusterIP 10.111.167.160 <none> 50051/TCP 36m service/clara-monitor-server-grafana NodePort 10.100.148.116 <none> 80:32000/TCP 36m service/clara-renderer-clara-render-server NodePort 10.108.60.232 <none> 8070:30105/TCP,80 60:32006/TCP 48m service/clara-resultsservice ClusterIP 10.109.206.204 <none> 8088/TCP 61m service/clara-ui ClusterIP 10.101.195.91 <none> 80/TCP 61m service/elasticsearch-master ClusterIP 10.108.240.18 <none> 9200/TCP,9300/TCP 36m service/elasticsearch-master-headless ClusterIP None <none> 9200/TCP,9300/TCP 36m service/kubernetes ClusterIP 10.96.0.1 <none> 443/TCP 65m NAME DESIRED CURRENT READY UP-TO-DATE AVAI LABLE NODE SELECTOR AGE daemonset.apps/clara-monitor-server-fluentd-elasticsearch 2 2 2 2 2 <none> 36m daemonset.apps/fluentd 2 2 2 2 2 <none> 61m NAME READY UP-TO-DATE AVAILABLE AGE deployment.apps/clara-clara-platformapiserver 1/1 1 1 61m deployment.apps/clara-console 1/1 1 1 19m deployment.apps/clara-console-mongodb 1/1 1 1 19m deployment.apps/clara-dicom-adapter 1/1 1 1 58m deployment.apps/clara-monitor-server-grafana 1/1 1 1 36m deployment.apps/clara-monitor-server-monitor-server 0/1 1 0 36m deployment.apps/clara-render-server-clara-renderer 0/1 1 0 48m deployment.apps/clara-resultsservice 1/1 1 1 61m deployment.apps/clara-ui 1/1 1 1 61m deployment.apps/clara-workflow-controller 1/1 1 1 61m NAME DESIRED CURRENT READY AGE replicaset.apps/clara-clara-platformapiserver-54c5c44bbd 1 1 1 61m replicaset.apps/clara-console-8565b4d565 1 1 1 19m replicaset.apps/clara-console-mongodb-85f8bd5f95 1 1 1 19m replicaset.apps/clara-dicom-adapter-7948fcd445 1 1 1 58m replicaset.apps/clara-monitor-server-grafana-5f874b974d 1 1 1 36m replicaset.apps/clara-monitor-server-monitor-server-59c8bf68f7 1 1 0 36m replicaset.apps/clara-render-server-clara-renderer-d79dd4779 1 1 0 48m replicaset.apps/clara-resultsservice-664477898f 1 1 1 61m replicaset.apps/clara-ui-6f89b97df8 1 1 1 61m replicaset.apps/clara-workflow-controller-69cbb55fc8 1 1 1 61m NAME READY AGE statefulset.apps/elasticsearch-master 2/2 36m

當然，如果像我這個白目的安裝就沒有那順利了…

錯誤嘗試：部署 Clara Platform 與 啟動 Helm Chart

在環境要求的部分，對我來說比較麻煩的是 Kubernetes 與 Helm 的版號，因為我的伺服器環境是與組員共用，所以一開始我決定保留同事需要的環境來硬幹，試試能不能安裝成功，如果真的不行再來嘗試退版安裝。

所以這段如果只是要完成 Deploy SDK 安裝的可以跳過，這邊只是因為我的一時興起所產生的錯誤紀錄而已，當然如果想看我怎麼焦頭爛額的可以繼續往下拉。

恩…我先跟大家說最後的嘗試結果好了，我最後還是退版了。不過我有將過程的一些錯誤紀錄保留下來，看看之後還有沒機會回來再看，絕對不是因為單純湊數字 XDDD

這邊接續安裝步驟-配置 NGC 憑證，在完成 platform chart 的下載後，試著啟動 platform，得到了第一條錯誤訊息：

1 2 3 4 5 6 7 8 9 10 11 12 13

$ clara platform start Error: could not find tiller Usage: platform start [flags] Flags: -h, --help help for start Global Flags: --config string config file (default is $HOME/.clara/config.yaml) --verbose verbose output Error: could not find tiller

在 Stack Overflow 上看到了一條類似錯誤訊息的提問，似乎重新初始化 Helm 即可：

1 2 3 4 5 6 7

$ helm init Error: unknown command "init" for "helm" Did you mean this? lint Run 'helm --help' for usage.

結果沒有 helm init！

查詢了一下 Helm 所找不到的 Tiller 到底是啥，根據 smalltown 所說，在 Helm2 中，Tiller 是用來安裝與管理其他應用服務的 K8S 元件，簡單來說 Tiller 是一個用來與 K8S API Server 溝通的 Service，不過由於權限設置與管理的問題，在 Helm3 的推出後就走向歷史了。

很不幸的，我的 Helm 是 v3 的版本：

1 2 3

$ helm version version.BuildInfo{Version:"v3.1.2", GitCommit:"d878d4d45863e42fd5cff6743294a11d28a9abce", GitTreeState:"$ lean", GoVersion:"go1.13.8"}

Helm2 與 Helm3 的變動已經屬於系統架構的變動，這個實在不好改。經過調查與論壇上發問，最後只好將 Helm 降版。我是透過 Binary Releases 安裝的方式，將版本降回到 v2.15.2。

降版後再次檢查 Helm 的版號，版號是正確了，但錯誤訊息依舊沒有消失：

1 2 3 4

$ helm version Client: &version.Version{SemVer:"v2.15.2", GitCommit:"8dce272473e5f2a7bf58ce79bb5c3691db54c96b", GitTreeS tate:"clean"} Error: could not find tiller

不過版本都降了，helm init 指令應該可以使用了：

1 2 3 4 5 6 7 8 9

$ helm init $HELM_HOME has been configured at /home/.helm. Tiller (the Helm server-side component) has been installed into your Kubernetes Cluster. Please note: by default, Tiller is deployed with an insecure 'allow unauthenticated users' policy. To prevent this, run `helm init` with the --tiller-tls-verify flag. For more information on securing your installation see: https://docs.helm.sh/using_helm/#securing-your-h$ lm-installation

再次檢查 Helm 的版號，可以發現多出了 Server，用 kubectl 查看正在運行的 Pod，可以看到 Tiller 正在努力工作：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21

$ helm version Client: &version.Version{SemVer:"v2.15.2", GitCommit:"8dce272473e5f2a7bf58ce79bb5c3691db54c96b", GitTree$ tate:"clean"} Server: &version.Version{SemVer:"v2.15.2", GitCommit:"8dce272473e5f2a7bf58ce79bb5c3691db54c96b", GitTree$ tate:"clean"} $ kubectl get pods --namespace kube-system NAME READY STATUS RESTARTS AGE coredns-6955765f44-bl6jd 1/1 Running 0 18h coredns-6955765f44-mtlxv 1/1 Running 0 18h etcd-esccluster-control-plane 1/1 Running 0 18h kindnet-dxlgj 1/1 Running 0 18h kindnet-hpvkw 1/1 Running 0 18h kindnet-qb5lm 1/1 Running 0 18h kube-apiserver-esccluster-control-plane 1/1 Running 0 18h kube-controller-manager-esccluster-control-plane 1/1 Running 0 18h kube-proxy-cvpmt 1/1 Running 0 18h kube-proxy-nspv9 1/1 Running 0 18h kube-proxy-trkh5 1/1 Running 0 18h kube-scheduler-esccluster-control-plane 1/1 Running 0 18h tiller-deploy-58f57c5787-bsfkh 1/1 Running 0 14m

好了，排除 Tiller 的錯誤訊息後，重新 platform 的 Chart 後再重新 start 一次，看看會不會成功。

1 2 3 4 5 6 7 8 9 10

$ clara pull platform ✔ Yes Clara Platform 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara Hint: use "clara platform start" or "clara platform restart" to deploy pulled Clara Platform. $ clara platform start Starting clara... RPC error: code = Unknown desc = namespaces "default" is forbidden: User "system:serviceaccount:kube-system:default" cannot get resource "namespaces" in API group "" in the namespace "default"

呃，又出現 rpc 的 error。看到錯誤訊息，幾個可能的猜測：

1. Server 安裝配置的問題：會考慮這個是因為我的 Tiller 後來降版後我自己重起的。
2. 權限問題：這個的機會比較大，在查資料的時候，幾乎碰到的是這個狀況。

算了先試試看 Helm2 文件中說的本地運行分 Tiller 試試看：

1 2 3 4 5 6

$ ~/bin/tiller [main] 2020/10/27 10:53:03 Starting Tiller v2.15.2 (tls=false) [main] 2020/10/27 10:53:03 GRPC listening on :44134 [main] 2020/10/27 10:53:03 Probes listening on :44135 [main] 2020/10/27 10:53:03 Storage driver is ConfigMap [main] 2020/10/27 10:53:03 Max history per release is 0

但文件中第二步驟連接到新的本地 Tiller 主機，看起來怪怪的，所還是沒做了，

直接放棄第一條路，先是試著處理權限問題好了，根據 Helm2 的 Role-based Access Control 說明與 GitHub 上的大神討論重新設定了連接，並 start platform：

1 2 3 4

$clara platform start Starting clara... NAME: clara Note: If there is a running instance of Clara Console, Clara Dicom Adapter or Clara Renderer, they should be restarted.

喔耶！ platform 起動後，就跟前面一樣來下載 Clara Deploy Services 的 Helm Charts：

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24

$ clara pull dicom ✔ Yes Clara Dicom Adapter 0.7.1-2008.1 Chart saved at: /home/.clara/charts/dicom-adapter Hint: use "clara dicom start" or "clara dicom restart" to deploy pulled Clara Dicom Adapter. $ clara pull render ✔ Yes Clara Renderer 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara-renderer Hint: use "clara render start" or "clara render restart" to deploy pulled Clara Renderer. $ clara pull monitor ✔ Yes Clara Monitor Server 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara-monitor-server Hint: use "clara monitor start" or "clara monitor restart" to deploy pulled Clara Monitor Server. $ clara pull console ✔ Yes Clara Management Console 0.7.1-2008.1 Chart saved at: /home/.clara/charts/clara-console Hint: use "clara console start" or "clara console restart" to deploy pulled Clara Management Console.

最後是心驚膽戰的最後一步：

1 2 3 4 5 6 7 8 9 10 11 12

$ clara dicom start Starting DICOM Adapter... NAME: clara-dicom-adapter $ clara render start NAME: clara-render-server $ clara monitor start Error: rpc error: code = Unknown desc = validation failed: [unable to recognize "": no matches for kind "PodSecurityPolicy" in version "extensions/v1beta1", unable to recognize "": no matches for kind "Deployment" in version "apps/v1beta2", unable to recognize "": no matches for kind "StatefulSet" in version "apps/v1beta1"] $ clara console start NAME: clara-console

就知道沒這麼好過年的，是有看到有個 Issue 在討論這問題的，不過必須承認的是，這個討論超過我這個初學者對於 K8S 的掌握了，我才剛入門沒幾天阿（崩潰

問了論壇的人得到的回覆還是要我降版 K8S，所以最終只能鼻子摸摸開始降版了：

1 2 3 4 5 6

$ sudo apt remove kubectl kubeadm kubelet kubernetes-cni $ rm -rf $HOME/.kube/config $ sudo apt-get install -y kubelet=1.15.4-00 kubectl=1.15.4-00 kubeadm=1.15.4-00 $ kubectl version --short Client Version: v1.15.4 Server Version: v1.15.6

重新啟動剛剛失敗 monitor Chart：

1 2	$ clara monitor start NAME: clara-monitor-server

參考資料

1. Clara Deploy Platform。檢自 NVIDIA NGC (2021-02-02)。
2. NVIDIA Taiwan (2020-11-02)。NVIDIA Clara Deploy。檢自 Youtube (2021-02-02)。
3. 2. Installation。檢自 Clara Deploy SDK https://hackmd.io/0.7.3 documentation (2021-02-02)。
4. Community (2019-05-18)。openshift - Helm: could not find tiller。檢自 Stack Overflow (2021-02-02)。
5. smalltown (2020-05-17)。Helm 3 踹踹看。檢自 Starbugs Weekly 星巴哥技術專欄｜Medium (2021-02-02)。
6. godleon (2021-01-24)。[Kubernetes] Package Manager - Helm 簡介。檢自 小信豬的原始部落 (2021-02-02)。
7. postak (2018-04-10)。forbidden: User “system:serviceaccount:kube-system:default” cannot get namespaces in the namespace “default。檢自 fnproject/fn-helm｜GitHub (2021-02-02)。
8. noprom (2017-11-12)。User “system:serviceaccount:kube-system:default” cannot get namespaces in the namespace “default”。檢自 helm/helm｜GitHub (2021-02-02)。
9. Helm v2。檢自 Helm 官網 (2021-02-02)。
10. Nick (2019-10-12)。[Day27] k8s應用篇（一）：Helm部署apps、HPA和CA。檢自 iT 邦幫忙 (2021-02-02)。
11. Terrones-Oscar (2020-08-13)。helm fails Error: validation failed: [unable to recognize “”: no matches for kind “PodSecurityPolicy”]。檢自 helm/charts｜GitHub (2021-02-02)。
12. jckasper (2019-09-06)。Helm init fails on Kubernetes 1.16.0。檢自 helm/helm｜GitHub (2021-02-02)。
13. Zz Chen (2018-07-03)。Helm 部署在 GKE 上的權限問題。檢自 smalltowntechblog｜Medium (2021-02-02)。
14. MengYun (2019-10-27)。Not responding when running “clara render start”。檢自 NVIDIA Developer Forums (2021-02-02)。

更新紀錄

最後更新日期：2021-03-07

• 2021-03-07 發布
• 2021-02-03 完稿
• 2020-11-09 起稿