ops(ci): upload full demo cluster log on Jepsen failure

The previous "Dump demo cluster log on failure" step ran
`tail -n 500 /tmp/elastickv-demo.log` which, for a 3-minute
workload, captures only the startup lines and discards the
window where the actual anomaly happens. Investigation of the
scheduled run 26198185540 (which surfaced a real
:duplicate-elements anomaly) was blocked because the GH UI
showed only server-startup logs, not the start_ts / commit_ts /
raft term / write-conflict / lock-resolver events that would
identify the offending code path.

Change:

  - Inline dump now prints head + tail (200 / 1000 lines for the
    single-process demo, 200 / 500 per node for the 3-node demo)
    so the GH UI still shows enough at-a-glance.
  - Full log uploaded as a 14-day artifact (`elastickv-demo-log`
    for the scheduled workflow, `elastickv-demo-logs` with three
    files for the per-push workflow). 14 days mirrors the
    existing Jepsen-store artifact retention so logs and
    history.txt can be correlated for the same run.

Self-review:

  1. Data loss — N/A.
  2. Concurrency — N/A.
  3. Performance — only fires on failure; artifact upload is a
     single tar-gz of a multi-MB log, negligible CI time.
  4. Data consistency — N/A.
  5. Test coverage — workflow change; no Go-side test surface.
     Verification = next failing scheduled run produces a
     downloadable artifact.

Author: bootjp

.github/workflows/jepsen-test-scheduled.yml

@@ -229,7 +229,28 @@ jobs:
             --host 127.0.0.1
       - name: Dump demo cluster log on failure
         if: failure()
-        run: tail -n 500 /tmp/elastickv-demo.log || true
+        # The previous `tail -n 500` truncated a 3-minute workload's
+        # log down to startup-only lines, making it impossible to
+        # correlate a Jepsen anomaly with the server-side state
+        # (start_ts, commit_ts, raft term, write conflicts, lock-
+        # resolver events). Print head + tail inline so the GH UI
+        # still shows the most recent activity at-a-glance, then
+        # upload the full log as an artifact for offline analysis.
+        run: |
+          echo "=== first 200 lines (startup) ==="
+          head -n 200 /tmp/elastickv-demo.log || true
+          echo "=== last 1000 lines (most recent activity) ==="
+          tail -n 1000 /tmp/elastickv-demo.log || true
+          echo "=== full log line count ==="
+          wc -l /tmp/elastickv-demo.log || true
+      - name: Upload demo cluster log on failure
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: elastickv-demo-log
+          path: /tmp/elastickv-demo.log
+          retention-days: 14
+          if-no-files-found: warn
       - name: Stop demo cluster
         if: always()
         run: |

.github/workflows/jepsen-test.yml

@@ -194,6 +194,33 @@ jobs:
             --partition-count 4 --group-count 6 \
             --drain-time 15 \
             --sqs-ports 63501,63502,63503 --host 127.0.0.1
+      - name: Dump demo cluster logs on failure
+        if: failure()
+        # Same rationale as jepsen-test-scheduled.yml: inline summary
+        # for the GH UI + full per-node logs uploaded as artifact so
+        # any Jepsen anomaly can be traced to the server-side ts/raft
+        # state without re-running the workflow.
+        run: |
+          for node in n1 n2 n3; do
+            log=/tmp/elastickv-demo-${node}.log
+            echo "=== ${node}: first 200 lines (startup) ==="
+            head -n 200 "$log" || true
+            echo "=== ${node}: last 500 lines (most recent activity) ==="
+            tail -n 500 "$log" || true
+            echo "=== ${node}: full log line count ==="
+            wc -l "$log" || true
+          done
+      - name: Upload demo cluster logs on failure
+        if: failure()
+        uses: actions/upload-artifact@v4
+        with:
+          name: elastickv-demo-logs
+          path: |
+            /tmp/elastickv-demo-n1.log
+            /tmp/elastickv-demo-n2.log
+            /tmp/elastickv-demo-n3.log
+          retention-days: 14
+          if-no-files-found: warn
       - name: Stop demo cluster
         if: always()
         run: |