Beta Release Guide

Stackdriver Kubernetes Engine Monitoring is now generally available. While we recommend that you upgrade to the GA version, GKE version 1.12.7, you can continue to use the Beta version of Stackdriver Kubernetes Engine Monitoring.

This guide provides details about the Beta releases of Stackdriver Kubernetes Engine Monitoring.

Release versions

Since the installation of Stackdriver Kubernetes Engine Monitoring is now managed by GKE, manual installation is no longer documented.

The current Beta releases and updates of Stackdriver Kubernetes Engine Monitoring are listed below with the final stable Beta version listed first:

  • Beta Update (Kubernetes 1.12.5): This update is strongly recommended for all users. To upgrade to this release, go to Installing Kubernetes Monitoring.

  • Other supported Beta versions (Kubernetes 1.10.12 / 1.11.6 / 1.11.7 / 1.11.8): To upgrade to any of these releases, go to Installing Kubernetes Monitoring.

    Before upgrading to any of these versions, be sure to check this page for known issues.

  • Original Beta (Kubernetes 1.10.2): Due to a number of known issues, this release is not recommended for any purpose.

Features and known issues

This section lists the known issues in the current Beta release versions. The following issues remain as of May 20, 2019:

Beta update 1.12.6

  1. LogEntry metadata label fields change

    The metadata label fields change when you upgrade your Google Kubernetes Engine clusters to GKE 1.12.6 or higher:

    • Kubernetes Pod labels, currently located in the metadata.userLabels field, are moved to the labels field in the LogEntry and the label keys have a prefix of k8s-pod/. The filter expressions in your sinks, logs-based metrics, exclusions, or queries might need to change.

    • Stackdriver system labels that are in the metadata.systemLabels field are no longer available.

  2. The metadata agent generates log spam

    The metadata agent has a known bug that generates log spam with a 400 Bad Request error. All metadata is still collected correctly, and features aren't impacted. However, this log spam consumes unnecessary log storage and may generate additional chargeable logs.

    Users should not upgrade to this release until this bug has been fixed.

    Fixed in 1.12.7-gke.10.

Beta update 1.11.6

To upgrade to this release, go to Installing Kubernetes Monitoring:

  • Includes major improvements to Stackdriver Kubernetes Engine Monitoring tabular UI loading and rendering latency.
  • Fixes most of the known issues with the 1.10.2 release.
  • Restores the ability to install the managed support for Stackdriver Kubernetes Engine Monitoring.
  • Lets you upgrade existing clusters with Stackdriver Kubernetes Engine Monitoring.
  • Removes the need for any manual installation on GKE.
  1. Stackdriver metadata agent restarts

    In the Beta release, the metadata agent can restart once per hour in normal operation, and might restart several times in succession.

    Fixed in 1.12.6-gke.7.

    Workaround: Do not worry about restart counts for the metadata agent.

Beta release 1.11.4

  1. Incorrect parsing of structured JSON logs as of Kubernetes 1.11.4

    Beginning with Kubernetes 1.11.4, structured JSON logs aren't parsed correctly. One effect is that alerts configured to trigger based on structured log content don't fire correctly. This can result in an alert failing to fire or firing when it shouldn't. Additionally, logs-based metrics that rely on structured log content are invalid.

    Workaround: Downgrade to GKE version 1.11.3:

    gcloud container clusters upgrade [CLUSTER_NAME] --zone [ZONE] --project [PROJECT] --master --cluster-version=1.11.3
    

    Fixed in 1.11.6-gke.2 and 1.12.4-gke.2.

Beta release 1.10.2

  1. Missing resource entries

    You might not see all your namespaces, workloads, pods, or other entities in Stackdriver Kubernetes Engine Monitoring. This is due to a problem in gathering metadata. It can take up to 30 minutes for the missing entities to be found. They appear after refreshing the display.

    Fixed in Kubernetes versions 1.10.6 and 1.11.2. You can also use the Manual Installation steps to install newer versions of Heapster and the Stackdriver agents in a new cluster.

  2. Regional and multi-zonal clusters

    Entities belonging to regional and multi-zonal clusters are not displayed correctly. The Infrastructure, Workload, and Service tabs do not expand to the node/pod level for these clusters.

    Fixed in Kubernetes versions 1.10.6 and 1.11.2. You can also use the Manual Installation steps to install newer versions of Heapster and the Stackdriver agents in a new cluster.

  3. Severity missing in log entries

    Log entries ingested by the logging agent are missing the severity field, which causes all log entries to show up with default severity (a gray star), breaking Error Reporting integration.

    Fixed in Kubernetes versions 1.10.6 and 1.11.2. You can also use the Manual Installation steps to install newer versions of Heapster and the Stackdriver agents in a new cluster.

  4. Exception stack trace detection is broken

    Multi-line exception stack traces are split into multiple separate log entries, breaking Error Reporting integration.

    Fixed in Kubernetes versions 1.10.6 and 1.11.2. You can also use the Manual Installation steps to install newer versions of Heapster and the Stackdriver agents in a new cluster.

  5. Irrelevant error messages from agents

    The metadata agent can write irrelevant error log messages after a previous instance of the agent shuts down incorrectly. The messages are:

    terminate called after throwing an instance of 'boost::exception_ ...'
    what():  Error listening on provided port.
    terminate called recursively
    Aborted (core dumped)
    

    Also, the logging agent will log Error calling Metadata Agent after the metadata agent terminates.

    Fixed in Kubernetes versions 1.10.6 and 1.11.2. You can also use the Manual Installation steps to install newer versions of Heapster and the Stackdriver agents in a new cluster.

  6. Stackdriver agents don't work behind Compute Engine firewalls

    Stackdriver Kubernetes Engine Monitoring does not work with clusters that are behind certain Compute Engine firewalls. In those situations, your metadata agent issues the following error message:

    Metadata request unsuccessful: Address family not supported by protocol.
    

    Fixed in Kubernetes versions 1.10.6 and 1.11.2. You can also use the Manual Installation steps to install newer versions of Heapster and the Stackdriver agents in a new cluster.

  7. Stackdriver metadata agent restarts

    In the Beta release, the metadata agent can restart once per hour in normal operation, and might restart several times in succession.

    Fixed in 1.12.6-gke.7.

    Workaround: Do not worry about restart counts for the metadata agent.

هل كانت هذه الصفحة مفيدة؟ يرجى تقييم أدائنا:

إرسال تعليقات حول...

Stackdriver Monitoring
هل تحتاج إلى مساعدة؟ انتقل إلى صفحة الدعم.