fix: roberta self-test was not base64-decoding slack secrets

Also not the correct channel field in the payload. This PR adds some helpers, resume.sh and suspend.sh, on the periodic task.
project-codeflare · Aug 30, 2022 · 3f3fad0 · 3f3fad0
1 parent 2c42713
commit 3f3fad0
Show file tree

Hide file tree

Showing 6 changed files with 39 additions and 25 deletions.
diff --git a/deploy/self-test/roberta/1gpu/once.yaml b/deploy/self-test/roberta/1gpu/once.yaml
@@ -49,32 +49,28 @@ spec:
       serviceAccountName: codeflare-self-test-serviceaccount
       containers:
       - name: self-test
-        image: ghcr.io/project-codeflare/codeflare-self-test:0.11.0
+        image: ghcr.io/project-codeflare/codeflare-self-test:0.11.1
         env:
           # - name: GUIDEBOOK_RUN_ARGS
           #  value: "-V"
           - name: VARIANTS
             value: roberta-1gpu
-          - name: SLACK_TEAMID
+          - name: SLACK_URL
             valueFrom:
               secretKeyRef:
                 name: codeflare-self-test-slack
-                key: TEAMID
-          - name: SLACK_CHANNELID
+                key: URL
+          - name: SLACK_CHANNEL
             valueFrom:
               secretKeyRef:
                 name: codeflare-self-test-slack
-                key: CHANNELID
-          - name: SLACK_TOKEN
-            valueFrom:
-              secretKeyRef:
-                name: codeflare-self-test-slack
-                key: TOKEN
+                key: CHANNEL
           - name: SLACK_USERNAME
             valueFrom:
               secretKeyRef:
                 name: codeflare-self-test-slack
                 key: USERNAME
+                optional: true
           - name: ML_CODEFLARE_ROBERTA_GITHUB_USER
             valueFrom:
               secretKeyRef:

diff --git a/deploy/self-test/roberta/1gpu/periodic.yaml b/deploy/self-test/roberta/1gpu/periodic.yaml
@@ -53,22 +53,22 @@ spec:
           serviceAccountName: codeflare-self-test-serviceaccount
           containers:
           - name: self-test
-            image: ghcr.io/project-codeflare/codeflare-self-test:0.11.0
+            image: ghcr.io/project-codeflare/codeflare-self-test:0.11.1
             env:
               # - name: GUIDEBOOK_RUN_ARGS
               #  value: "-V"
               - name: VARIANTS
                 value: roberta-1gpu
-              - name: SLACK_TEAMID
+              - name: SLACK_URL
                 valueFrom:
                   secretKeyRef:
                     name: codeflare-self-test-slack
-                    key: TEAMID
-              - name: SLACK_CHANNELID
+                    key: URL
+              - name: SLACK_CHANNEL
                 valueFrom:
                   secretKeyRef:
                     name: codeflare-self-test-slack
-                    key: CHANNELID
+                    key: CHANNEL
               - name: SLACK_TOKEN
                 valueFrom:
                   secretKeyRef:
@@ -79,6 +79,7 @@ spec:
                   secretKeyRef:
                     name: codeflare-self-test-slack
                     key: USERNAME
+                    optional: true
               - name: ML_CODEFLARE_ROBERTA_GITHUB_USER
                 valueFrom:
                   secretKeyRef:

diff --git a/deploy/self-test/roberta/1gpu/resume.sh b/deploy/self-test/roberta/1gpu/resume.sh
@@ -0,0 +1,5 @@
+#
+# Resumes the `periodic` cronjob.
+#
+
+kubectl patch cronjobs codeflare-self-test-roberta-1gpu-periodic -p '{"spec" : {"suspend" : false }}'
diff --git a/deploy/self-test/roberta/1gpu/suspend.sh b/deploy/self-test/roberta/1gpu/suspend.sh
@@ -0,0 +1,6 @@
+#
+# Suspends the `periodic` cronjob. This is useful if you want to sneak
+# in a `once` run for debugging.
+#
+
+kubectl patch cronjobs codeflare-self-test-roberta-1gpu-periodic -p '{"spec" : {"suspend" : true }}'
diff --git a/deploy/self-test/self-test.sh b/deploy/self-test/self-test.sh
@@ -20,12 +20,7 @@ echo "Using these variants: $variants"
 function cleanup {
     STATUS=$?
 
-    TEAMID=$SLACK_TEAMID
-    CHANNELID=$SLACK_CHANNELID
-    TOKEN=$SLACK_TOKEN
-    USERNAME=${SLACK_USERNAME-"Self-test Bot"}
-
-    if [ -n "$TEAMID" ] && [ -n "$CHANNELID" ] && [ -n "$TOKEN" ]; then
+    if [ -n "$SLACK_URL" ] && [ -n "$SLACK_CHANNEL" ]; then
         if [ "$STATUS" = "0" ]; then
             ICON=":green_heart:"
             TEXT="$(date)\nSuccessful run\nVariants: $variants"
@@ -35,10 +30,21 @@ function cleanup {
             # TODO, add tee'd logs of failing run
         fi
 
+        URL=$(echo $SLACK_URL | base64 -d)
+        CHANNEL=$(echo $SLACK_CHANNEL | base64 -d)
+
+        if [ -n "$SLACK_USERNAME" ]; then
+           USERNAME=$(echo $SLACK_USERNAME | base64 -d)
+        else
+            USERNAME="Self-test Bot"
+        fi
 
+        echo "Posting to slack status=$STATUS to $URL"
         curl -X POST \
              --data-urlencode "payload={\"channel\": \"#$CHANNEL\", \"username\": \"$USERNAME\", \"text\": \"$TEXT\", \"icon_emoji\": \"$ICON\"}" \
-             https://hooks.slack.com/services/$TEAMID/$CHANNELID/$TOKEN
+             $URL
+    else
+        echo "Skipping post to slack url?=${#SLACK_URL} channel?=${#SLACK_CHANNEL}"
     fi
 }
 

diff --git a/docs/self-test/roberta/README.md b/docs/self-test/roberta/README.md
@@ -57,9 +57,9 @@ export SLACK_TOKEN=mytoken
 
 ```shell
 kubectl create secret generic codeflare-self-test-slack \
-    --from-literal=TEAMID=$(echo -n SLACK_TEAMID | base64) \
-    --from-literal=CHANNELID=$(echo -n SLACK_CHANNELID | base64) \
-    --from-literal=TOKEN=$(echo -n SLACK_TOKEN | base64)
+    --from-literal=TEAMID=$(echo -n $SLACK_TEAMID | base64) \
+    --from-literal=CHANNELID=$(echo -n $SLACK_CHANNELID | base64) \
+    --from-literal=TOKEN=$(echo -n $SLACK_TOKEN | base64)
 ```
 
 ### Deploy the automation