Jupyter kernel disconnect kernel dies under memory pressure

By Sai Kiran Pandrala · Last verified: 2026-05-31 · Source: Google Cloud Community, community Q&A, Google Cloud docs

Engineers running Vertex AI Workbench hit Jupyter kernel disconnect kernel dies under memory pressure often enough that there is a stable fix pattern. This page captures it in the order Google Cloud support would run it during a real incident.

What jupyter kernel disconnect kernel dies under memory pressure actually involves on Vertex AI Workbench

This task on Vertex AI Workbench is one of the more searched operational topics on AWS in the last 12 months. The procedure below is the path that works in a current AWS account with default IAM and standard VPC config.

The rest of this page is the structured fix path. Start with diagnose, then remediation, then the automation options so you do not have to do this by hand the next time it surfaces. Verify and safety sections at the end are the discipline that keeps the fix from regressing in production.

Diagnose first, fix second

Run gcloud auth list and gcloud config list first. About one in five 'why does this not work' tickets are actually 'I am in the wrong account' or 'my session expired and the SDK is using stale credentials or ADC pointed at the wrong project'. The 5-second sanity check costs nothing and saves real time when the answer is that simple.

Check Cloud Monitoring Logs for the calling service. Lambda, ECS, EKS, Step Functions, API Gateway, and most managed services write detailed traces to Cloud Monitoring Logs under predictable log group names. Use Cloud Monitoring Logs Insights with fields @timestamp, @message | filter @message like /ERROR/ | sort @timestamp desc | limit 50 to surface the most recent failures.

Start by capturing the exact Google Cloud error string. The Cloud Console truncates messages in popups, but Cloud Logging keeps the full record in protoPayload.status and protoPayload.methodName. The camelCase error code (e.g. AccessDenied, InsufficientInstanceCapacity, ConditionalCheckFailedException) is the thing you grep for in Google Cloud Community and StackOverflow, not the human-readable sentence next to it. Paste the code into the re:Post search bar in quotes and you will usually land on at least one Google-staff-verified answer within the first three results.

Solution-focused remediation path

Most Vertex AI Workbench failures fall into one of three buckets: IAM permission gap, networking path break (security group, NACL, or VPC endpoint policy), or service-limit / quota hit. Run that mental triage first - it covers around 80 percent of real-world cases. If the failure does not fit any of the three, it is likely a service-side regression worth opening a re:Post or support ticket for.

If networking is suspect, use Network Intelligence Connectivity Tests. It is the only tool that simulates the full ENI-to-ENI path including firewall rules, hierarchical firewall policies, routes, and VPC Service Controls perimeters in one call. Manual trace is slower and misses transitive issues. The analyzer charges $0.10 per analysis - cheaper than a 30-minute call with your network team.

If the issue points at IAM, do not start by adding * to a policy. Use IAM Policy Troubleshooter and IAM Recommender against the failed action to see the minimum scope. Adding * is the fastest way to fail your next Google Cloud Architecture Framework security review, and it usually does not even fix the issue because the explicit deny is often coming from a higher level (Org Policy, RCP, or permission boundary), not a missing allow.

Automate this fix so you do not do it twice

Add a Workflows or Cloud Tasks Automation runbook

For multi-step fixes that include a manual approval, use Workflows runbook. Document the fix as a runbook with workflows.executions.approve steps where a human signs off and workflows.steps.callApi steps where the runbook calls the Google Cloud API. Approvers are notified by SNS; the runbook execution shows up in Cloud Audit Logs with the approver's identity attached. This makes audit trails easy and stops production fixes from being one-person operations.

Wire the fix into Eventarc for self-healing

If the failure mode is recurring, automate the remediation instead of the diagnosis. Eventarc Scheduler or rules that watch Cloud Logging events for the specific error code can invoke a Lambda that runs the same fix you would run by hand. The Lambda must be idempotent (re-running it on already-healthy resources must be a no-op) and must emit a Cloud Monitoring metric so you can track how often the auto-fix fires. A spike in auto-fix invocations is itself a signal worth alerting on.

# Eventarc rule pattern (JSON)
{ "source": ["aws.vertex"], "detail-type": ["Google Cloud API Call via Cloud Audit Logs"], "detail": { "errorCode": ["AccessDenied", "ThrottlingException"] }
}

Add a Cloud Monitoring alert policy so you know next time

The cheapest way to never see the same incident twice is a Cloud Monitoring alert policy on the metric that would have warned you. For Vertex AI Workbench, the relevant metrics live under compute.googleapis.com/vertex namespace or under custom metrics published by your Cloud Run service or GKE pod. Set thresholds based on observed normal range plus one or two standard deviations, not on round-number guesses. Cloud Monitoring anomaly-based alert policies remove the threshold-guessing problem entirely for metrics with regular seasonality.

Common pitfalls and what to watch for

The most common pitfall when fixing this on Vertex AI Workbench is treating it as a one-off rather than as a recurring class of incident. The same misconfiguration tends to happen again after a deployment, a role rotation, or a region migration unless the fix is codified. Add a Org Policy or VPC Service Controls constraint, Organization Policy condition, or Org Policy or VPC Service Controls rule that prevents the same misconfig from being introduced again. Documentation alone does not survive turnover.

Another common trap: confirming the fix on a single resource and assuming the fleet is healthy. Loop your check across every account, region, and IAM principal that could exhibit the same symptom. If you cannot enumerate the affected scope without a script, you do not yet understand the scope.

Verify the fix worked

• Reproduce the original symptom path. If it still surfaces in any account or region or IAM role or service account, you have not fixed it.
• Watch for 24 to 48 hours. Cloud Monitoring metrics and Cloud Asset Inventory can mask issues with cached health for 6 to 12 hours, especially Cloud CDN and Cloud DNS.
• Run a smoke test under realistic load. Happy-path tests miss race conditions and IAM session-cache issues.
• Capture the new state in a runbook so the next person on call does not have to rediscover this. Push it to Confluence or your team wiki, not into Slack.
• If the fix involved a permission change, run IAM Access Analyzer one more time to confirm you did not open a separate hole while closing this one.

Safety, rollback, blast radius

• Test in a non-production account if your environment has Resource Manager and Organization Policy or Cloud Resource Manager (organizations, folders, projects). The cost of one sandbox account is cheaper than one rollback meeting.
• Export the existing config before changing it. Most Vertex AI Workbench resources support describe + export to JSON via CLI - capture that to source control before you start.
• Know your rollback path. Some Vertex AI Workbench operations are one-way (region migration, account-level feature opt-in, Cloud KMS key deletion past pending window). Confirm reversibility on the Google Cloud doc before you commit.
• Be aware of cross-service impact. IAM role or service account changes ripple to every service trusting that role. Cloud KMS key changes break every workload depending on that key. VPC endpoint changes affect every VPC consumer of that endpoint.
• Maintenance window discipline: if the change touches DNS, certificate rotation, or anything that emits TLS handshakes, line up a window with stakeholder notification, not a heroic mid-day swap.

FAQ

Related fixes

Related guides worth a look while you sort this one out:

• Workbench instance fails health check after kernel upgrade
• Online endpoint 503 UNAVAILABLE under autoscale ramp
• Conda env conflicting dependencies after pip install in managed notebook
• Configure single-user vs service-account auth mode
• Cost optimization stop vs delete vs idle-shutdown for Workbench
• Custom container image fails to pull on Workbench instance