Elastic Cloud Enterprise 2.10.0edit

Release highlightsedit

The frozen data tier is now available in Elastic Cloud Enterprise. A frozen data tier, powered by searchable snapshots, allows you to store large volumes of less frequently searched data in object storage including Amazon S3, Google Cloud Storage, and Microsoft Azure Blob Storage. The new data tier is designed to be so cost-effective you’ll never need to delete data. It is fully integrated with autoscaling, allowing you to seamlessly scale to store massive amounts of data. Check Frozen tier.

New custom endpoint aliases. Custom endpoint aliases allow you to change the overly long, robotic URLs for your deployment and Elastic applications to something you choose. These predictable, human-readable URLs can be shared easily. An alias is unique to only one deployment within a region. They don’t replace the existing URL, but work alongside it. Best of both worlds. Check Custom endpoint aliases.


Display CPU when creating a deployment. The amount of CPU for each instance is now displayed when creating a deployment.

Add instrumentation to ZooKeeper create calls. Adds instrumentation to ZooKeeper create calls, in order to track write transaction count and rate for the service.

Add data tier instance filter. Adds the ability to filter deployment instances by data tier on the deployment overview page.

APM plan handler to wait for Kibana version if APM version supports Fleet server. While processing an APM plan that supports Fleet Server (7.13+) the ECE constructor now waits for a Kibana cluster to be the same or greater version. This means that during upgrade of a deployment to version 7.13+, the constructor will not proceed with an APM plan until the Elasticsearch and Kibana plans are finished.

Frozen Tier stats display. Changes the display of frozen data nodes on the overview page. We were previously displaying normal disk allocation values, but those aren’t applicable because of searchable object storage. We now display the usage of that searchable object storage and the actual node size.

Add Kibana configuration for cgroup based monitoring. With this change, the Stack Monitoring UI shows CPU and memory resource usage based on collected cgroup values for APM Server in versions 7.13+.

Show a warning on deployments using deprecated cross-cluster search templates. Owners of deprecated cross-cluster search deployments are now prompted to migrate to a new template prior to upgrading to future Elastic Stack versions.

Wait until API is healthy to exit upgrade process. This adds a step in the ECE upgrader to wait until the API is ready before exiting the process.

Platform UI redesign. Redesigned the UI around hosts, role management, allocators, proxies, and control planes.

Increase ZooKeeper timeouts. - Doubles the ZooKeeper initLimit and syncLimit timeouts to 14 and 6 respectively. Applies to ZooKeeper server processes and observers.

Move hardware profile to dropdown in advanced settings. When creating a deployment, the different hardware profile options are now available on the Settings menu.

Do retries for failed keystore writes. Make allocations more resilient to transient Docker API errors by retrying Docker keystore writes if they fail.

Account UI Amsterdam fixes. When updating to the new Amsterdam EUI theme, it became clear that there were some visual issues in the account section. This solves many of them.

Dismiss autoscaling limit reached banner. Adds the ability to dismiss the "Autoscaling limits reached" banner.

Frozen tier display blob storage. Displays frozen tier storage as the amount of searchable storage available (a multiplier of 1600x the RAM size).

Display changes in activity view by topology element ID if present. Configuration changes will now reference Elasticsearch topology elements by ID instead of their underlying hardware configuration, where possible.

Fix styling for allocator query builder. The styling for the allocator query builder has been improved.

Update SLM health status behavior. When snapshot lifecycle management (SLM) is in use, snapshots should only be reported as unhealthy if SLM is suspended while there is no ongoing plan activity.

Only ignore allocation.require shard routing rules. Ensure that data tier shard routing allocation preferences are retained when restoring a deployment from snapshots, or restoring a snapshot from another deployment.

Upgrade ECE logging and metrics cluster bootstrap. As of version 2.9, ECE now automatically upgrades the Logging and Metrics system cluster to the latest bundled 7.x version of Elasticsearch. All new ECE installations 2.9 and later will also be created with a 7.x Logging and Metrics cluster.

Update UI to suppoort data tier removal. The updated UI enables users to disable optional data tiers.

System cluster create and update in parallel. The ECE installer now creates the Security and Logging & Metrics clusters at the same time, speeding up the initial installation time by several minutes. The ECE upgrader also now upgrades all eligible system clusters at the same time.

Reference certificate for default domain. Previously, we assumed the primary SSL certificate would have the root domain of found.io. This change makes it so we don’t assume a default domain at all. This was done in order to avoid a potentially very slow path of checking extraCerts. With recent work to improve the speed of extraCerts this is no longer necessary.

Remove NodeAttributeValidation from plan prerequisite validation. Remove plan validation which rejects plans that remove node attributes. This allows users to remove optional data tiers through the API.

Bug fixesedit

Convert restore snapshot to deployment API. Fixed an issue where restoring snapshots on deployments that have converted to node_roles format was not working.

Add subdomain for new deployment alias. Fixed a bug where the application subdomain, such as es for Elasticsearch and kb for Kibana, were not appearing in the URL displayed on the custom endpoint alias UI for new deployment aliases.

Do not send autoscaling values if autoscaling is not turned on. Removes autoscaling properties from being sent on create and update requests when autoscaling_enabled is set to false. This creates a better experience for users who view the equivalent API request, or if restoring from a snapshot.

Remove empty observability from plans. Fixes an issue where disabling logs and metrics would include an empty observability section in the plan change instead of removing the field entirely.

Return error on missing resources requiring SSO. A bug in Cloud SSO has been fixed where Deployment API PUT payloads are accepted that were missing resources requiring SSO SAML configuration. In this scenario the API call would succeed but the resulting plan change results in SSO breaking for the missing resources. Instead, an error is now returned for each missing resource requiring SAML configuration changes.

Support implicit stateless cluster versions in observability code. Fixed the following two bugs:

  • When using implicit version in stateless resources (Kibana, APM, and Enterprise Search) the Logs and Metrics page doesn’t show links to logs and metrics.
  • When using implicit version in stateless resources (Kibana, APM, and Enterprise Search) legacy monitoring may be used instead of Metricbeat monitoring.

Fix calculated timeout value. Fixes a bug where the calculated timeout for a deployment’s shutdown operation was too small, incorrectly causing shutdown plans to fail.

Show coordinating tier info in architecture summary. Fixed a bug that caused the summary section of the deployment page to not display settings for the coordinating tier.

SAML UI: Avoid initializing the use_single_logout option when the form is loaded with value false. Fixed a bug in ECE UI which made SAML Use single logout (SLO) checkbox to be initially rendered as checked even when that setting is actually disabled in the configuration.

Fixed warning when logs and metrics are disabled. Fixed a false negative warning "Resource plans are inconsistent with deployment logging and metrics settings. Please reapply desired settings" that was shown after disabling deployment logs and metrics.

Remove 0 size option for warm and cold tiers. Removes the ability to set a 0 size on warm and cold tiers if they are already sized when autoscaling is turned on. Removing warm and cold tiers isn’t currently supported as a safe operation.

Remove padding from Reset credentials modal. There was a minor UI bug that led to the Elastic credentials showing without left padding in the Reset credentials modal window. This has now been fixed.

More selective voting exclusions. Fixed a bug that would add voting exclusions to Elasticsearch master nodes unnecessarily during forced restarts, sometimes leading to plan failures later.

Use wait_for_removal=false when cleaning up voting exclusions at the end of a plan. Plans for 7.x Elasticsearch clusters will now clean up lingering voting exclusions more consistently to prevent future plans from failing.

Logs and metrics - deployment query adds check for owner_id. Fixed a bug in the Logs and metrics UI where for certain users a deployment is displayed from separate accounts when selecting a deployment.

Trust management section loading after a trusted deployment is deleted. Fixed a UI bug on the Trust management page that resulted in a permanent loading state once a trusted deployment was deleted.

Remove thumbnail disabling pre-7.11.0. Removed the availability of workplace_search.content_source.sync.thumbnails.enabled configuration in 7.10.2, as this option is not available in the version 7.10.x line.

Fix instance health reporting. Fixed a bug that would sometimes cause instances to be reported as healthy even if they were not. This could lead to cluster unavailability since these "unhealthy" instances would be taken out of the load balancer.

Fix App Search to Enterprise Search upgrades. Fixed a bug that would prevent users from migrating App Search deployments to Enterprise Search for versions >= 7.10.0.

Add delay between cluster health checks. Fixed a bug that would cause the wait-until-cluster-ready-for-rolling constructor step to very aggressively check the cluster’s health (and issue an enormous amount of writes to Zookeeper) under certain conditions.

Increase size of Training button. Ensures that the size of the Training button matches the size of other, similar buttons on the Elastic Cloud portal page.

Only ignore allocation.require shard routing rules. Ensure that data tier shard routing allocation preferences are retained when restoring a deployment from snapshots, or restoring a snapshot from another deployment.

Use aliased endpoints for logs and metrics links. Fixed an issue where the links to logs and metrics single-sign on doesn’t work.

Check node_roles for voting_only designation. Fixed a bug that could lead to quorum loss for Elasticsearch clusters version 7.11 or higher that are in exactly 2 availability zones (plus a tiebreaker) when running a plan to switch to dedicated masters.

Stop trying to query non-existent App Search. The Deployment Overview tab should no longer display an error if App Search is available but not enabled.

Adjust region query in the UI. Adjusted the UI query for selecting deployments within the same region to be less flexible.

TCP RST cleanup. Fixed a bug where an improperly closed connection to the transport passthrough proxy would result in an idle backend connection remaining open.

Don’t rollback start-instances or allocate-instances. Instances created during a plan that fails will no longer attempt to be removed as part of the rollback (except when standing up brand new clusters). Deleting instances as part of rollback is prone to failure (making things even worse), and often isn’t the best choice anyway.

Fix crashes in the deployment template view when the monitoring deployment couldn’t be found. When a deployment template included observability configuration, the deployment template view would incorrectly search for the referenced deployment by cluster ID, rather than deployment ID. When this failed to match any deployments the entire page would fail. We now correctly search by deployment ID, and show a relevant error when no deployment is found.

Fix displayed frozen storage values. Displays the proper values for frozen storage in the summary footer on the edit page of the UI, and in the architecture summary bar.

Filter cluster metadata during ConfigureAutoExpandIndexes. The plan step "Adjusting auto-expand indexes in Elasticsearch" no longer queries metadata for all indices in the cluster. This change makes this step much faster for large clusters, reduces load imposed on the cluster, and prevents it from failing for clusters with several hundred megabytes of cluster state data.

Filter out searchable snapshots on SLM deployments only. Filters out searchable snapshots in snapshot lists for SLM deployments only, as they are incomplete snapshots that should not be restored from.

Fix corrupt OOM heap dumps. Fixed an issue where heap dumps are sometimes corrupt when uploaded to snapshot storage.

ECE platform settings fixes. Fixed a few title size issues on the platform settings page of the UI, as well as an alignment problem for the the switch button.


Removed security and monitoring selections from deployment creation. During the deployment create workflow, the monitoring and security options have been removed. They can still be set on their respective pages after the deployment has been spun up.

Known issuesedit

Fleet configuration not working in ECE 2.10 with stack version 7.14. In ECE version 2.10 with the original Elastic Stack pack version 7.14, if downloaded from the Elastic website before August 10, 2021, Fleet does not work when enabled in a deployment. To support Fleet, you can get and re-upload a fresh copy of the version 7.14 Elastic Stack pack to overwrite the original one. If you have existing version 7.14 deployments, then restart Fleet/APM after re-uploading the Elastic Stack pack to enable Fleet. This issue will be addressed in later stack packs and ECE versions.