v5.5.0

Row-level access control, ip and uuid datatypes, column alias management in UI

Notable new features⚓︎

Row-level access control⚓︎

Data can now be filtered for users based on the contents of the row and users' assigned roles.
Each filter is implemented as a row policy.
RBAC roles control which users are affected by which row policies.
Learn more in Row-Level Access Control

Column alias management for data tables⚓︎

You can now create and manage column aliases directly from the UI. Tabs are fully synchronized to preserve input when switching.

Column alias list
In the UI data/tables/{table_id}/column-management/: View existing aliases with columns for Alias and Expression
Create new alias
In the UI /data/tables/{table_id}/column-management/new-column-alias/
- Alias form: Add a name and expression
- Analyze expression:
- Select optional columns to preview with the alias
- See generated SQL, copy it, and set row limits
- Run a test query to preview results
Edit alias
In the UI /data/tables/{table_id}/column-management/{alias_id}/: Edit or delete existing aliases

Support for new `ip` and `uuid` datatypes⚓︎

IP addresses and UUIDs may now be used as a first-class datatype in transforms, ingest, and queries.

The ip datatype supports both IPv4 and IPv6 addresses. Here's an example of field definitions in a transform:

{
  "name": "ipv4_field",
  "datatype": {
    "type": "ip",
    "default": "127.0.0.1"
  }
}

{
  "name": "ipv6_field",
  "datatype": {
    "type": "ip",
    "default": "fe80::22bb:2ebd:da90:3439"
  }
}

- The uuid datatype supports standard 128-bit UUIDs. Here's an example of a field definition in a transform:

{
  "name": "uuid_field",
  "datatype": {
    "type": "uuid",
    "default": "88a7bdff-ce1f-4ce6-ae6f-f506937e315c"
  }
}

Upgrade instructions⚓︎

⚠️ Do not skip release versions.

Skipping versions during upgrades may result in system instability. You must upgrade sequentially through each release version.

Example: Upgrade from 5.4 → 5.5 → 5.6, not 5.4 → 5.6

Apply the new Hydrolix operator

If you have a self-managed installation, apply the new operator directly with the kubectl command examples below. If you're using Hydrolix-supplied tools to manage your installation, follow the procedure prescribed by those tools.

### GKE

kubectl apply -f "https://www.hydrolix.io/operator/v5.5.0/operator-resources?namespace=${HDX_KUBERNETES_NAMESPACE}&gcp-storage-sa=${GCP_STORAGE_SA}"

### EKS

kubectl apply -f "https://www.hydrolix.io/operator/v5.5.0/operator-resources?namespace=${HDX_KUBERNETES_NAMESPACE}&aws-storage-role=${AWS_STORAGE_ROLE}"

### LKE and AKS

kubectl apply -f "https://www.hydrolix.io/operator/v5.5.0/operator-resources?namespace=${HDX_KUBERNETES_NAMESPACE}"

Monitor the upgrade process

Kubernetes jobs named init-cluster and init-turbine-api will automatically run to upgrade your entire installation to match the new operator's version number. This will take a few minutes, during which time you can observe your pods' restarts with your favorite Kubernetes monitor tool.

Ensure both the init-cluster and init-turbine-api jobs have completed successfully and that the turbine-api pod has restarted without errors. After that, view the UI and use the API of your new installation as a final check.

If the turbine-api pod doesn't restart properly or other functionality is missing, check the logs of the init-cluster and init-turbine-api jobs for details about failures. This can be done using the k9s utility or with the kubectl command:

% kubectl logs -l app=init-cluster
% kubectl logs -l app=init-turbine-api

If you still need help, contact Hydrolix support.

Changelog⚓︎

Updates⚓︎

These changes include version upgrades and internal dependency bumps.

Cluster Operations⚓︎

Upgraded gunicorn from 20.0.4 -> 20.1.0 to prevent cluster problems. See more information about the bugfix here: https://github.com/benoitc/gunicorn/issues/1913
Upgraded kopf and update operator RBAC from v0.0.0-20220102113305-2298ace6d09d to v0.0.0-20250630131328-58d95d85e994.
Upgraded HTTP Proxy to version 0.4.0.
Switched from pycryptodome to pycryptodomex version 3.23.0 to eliminate a security vulnerability.

Intake and Merge⚓︎

Updated Goja dependency from v0.0.0-20220102113305-2298ace6d09d to v0.0.0-20250630131328-58d95d85e994.
Updated the following crates: fake to version 4.3.0 and rand to version 0.9.1
Upgraded the golang.org/x/oauth2 library from version 0.23 to 0.27 to resolve a high-severity vulnerability (CVE-2025-22868). The issue allowed attackers to send malformed tokens that could cause excessive memory consumption during parsing.
Related advisory: GHSA-6v2p-p543-phr9
Updated go runtime from version 1.23.0 to 1.23.3 to address a segfault during autoingest.

UI⚓︎

Upgraded form-data library to Upgrades 2.5.4, 3.0.4, 4.0.4 or above to prevent Math.random()-related security vulnerability.

Improvements⚓︎

Config API⚓︎

Added /download endpoint to DictionaryFile resource that downloads the dictionary file using the same permissions as view_dictionaryfile
Implemented service account token generation within the config API. This allows more fine-grained control over token generation, TTLs, and revocations.
Updated the /defaults Config API endpoint to return updated default settings values. Updated schema for multiple endpoints for more clarity on the 'settings' field. Changed the batch job default max_minutes_per_partitions from 360 to 60.
Support for new ip and uuid datatypes has been added. See Support for new ip and uuid datatypes for more information.
Enabled storing HDX Deployment ID (deployment_id) on multi-tenant Project resources.
API query string parameter values for the /project endpoint are no longer converted to lowercase, preserving case-sensitivity.

Cluster Operations⚓︎

Hydrolix now includes both cluster-level and project-level deployment IDs in all usage reports, logs, and monitoring data.
This ensures traceability across usage.meter, hydro.logs, and alerting pipelines.
Deployment ID sources:
- HDX_DEPLOYMENT_ID: cluster-wide, env var or CLI flag
- project.hdx_deployment_id: per project, config data
Improved initialization logic to ensure the hydro.logs transform includes the deployment_id column, allowing logs to be filtered by deployment.
Added a service to guarantee a transform contains specific output_columns.
Updated init_ci to use this service to ensure deployment_id is present.
Refactored transform business logic into TransformSerializer for reuse and clarity.
Creates or updates the transform and settings only if deployment_id is missing.
Leaves other transform settings untouched.
Raises clear errors if a conflicting datatype for the column exists.
Include Hydrolix deployment ID in logs and metrics emitted by a cluster. The deployment ID is propogated to various workloads by the operator. If the operator can't obtain a deployment ID, it defaults to propagating the namespace.
When Grafana is enabled in-cluster, the operator creates a default Hydrolix datasource in Grafana, replacing the default ClickHouse datasource.
Added two metrics: fqdn_cert_expiry, which indicates the fqdn cert expiry time in seconds, and fqdn_tls_cert_expiry_diff, which indicates if the fqdn cert and the tls.crt in traefik-tls secret differ (which should not be the case).
Introduced multiple enhancements to the hdx-scaler autoscaler to support more responsive and stable scaling:
Separate cooldowns: cool_up_seconds and cool_down_seconds now control scale-up and scale-down cooldowns independently.
Range-based scaling: metric_min and metric_max define bounds for a metric; autoscaling adjusts pod count relative to that range.
Scaling throttles: New scale_up_throttle and scale_down_throttle options limit how quickly scale adjustments are made (as a percentage).
Tolerance window: tolerance_up and tolerance_down define a dead zone around the target where no scaling is triggered.
Added an interactive help window to the hdx-scaler terminal UI. Improves usability for users managing scaling settings from the TUI.
Press h to open the help panel, which explains available configuration fields.
Press ESC or q to close it.
The hdx-scaler service now exposes detailed Prometheus metrics to improve the observability of autoscaling behavior.

These metrics include:

hdxscaler_total_scalers: The total number of autoscalers running
hdxscaler_ratio: The calculated ratio based on abs(normalized metric - normalized target) if using Metric Range mode or target metric / observed metric if using Metric Target mode
hdxscaler_current_replicas: The current number of replicas
hdxscaler_target_count: The number of targets the measured value is being averaged from
hdxscaler_measured_value: The measured average value of the target group metrics
hdxscaler_desired_replicas: The calculated number of replicas to scale to before applying any bounds checks, sensitizing, or throttling
hdxscaler_bounded_replicas: What desired replicas are set to after applying all bounds checks, sensitizing, and throttling
hdxscaler_normalized_metric: What the measured value is normalized to if using Metric Range as the scaling mode
hdxscaler_normalized_target: What the target value is normalized to if using Metric Range as the scaling mode
hdxscaler_delta: The difference between the normalized metric and normalized target, used in Metric Range mode
hdxscaler_sensitivity_calc: The value calculated by the sensitizer, used in Metric Range scaling mode to calculated the desired replicas
hdxscaler_throttle_up_calc: If scaling up, the max number of pods to scale up to, will be used instead of desired replicas if smaller
hdxscaler_throttle_down_calc: If scaling down, the number of pods to scale down by, will be used if larger than desired replicas

Each metric (except hdxscaler_total_scalers) has the following labels:

deployment: the Deployment being scaled
alias: if that deployment has an alias, i.e. intake-head, alias="stream-head"
app: the service or pool name being scraped for metrics to watch
title: The name for this autoscaler which includes deployment name and metric / metric labels that are being used for scale calculations
HDX Scaler now supports loading its config from a Kubernetes ConfigMap using the CONFIG_MODE=k8s environment variable.
Faster config changes
Default remains unchanged
Improves deployment stability
HDX Scaler now ignores services when the scale_off:true tunable is set. This prevents accidental restarts of pods that should be scaled off.
To fully shut down the cluster, HDX Scaler must be scaled to zero manually.
HDX Scaler now supports loading its config from a Kubernetes ConfigMap using the CONFIG_MODE=k8s environment variable.
Faster config changes
Default remains unchanged
Improves deployment stability
HDX Scaler now ignores services when the scale_off:true tunable is set. This prevents accidental restarts of pods that should be scaled off.
To fully shut down the cluster, HDX Scaler must be scaled to zero manually.
Added regular expressions (regex) for header and query-based dynamic ingest routing. To enable regex matching, prefix the value with regex|. See the example below:

routing:
  headers:
    x-hdx-table: regex|staging-project-(alpha|beta|gamma)
    x-hdx-transform: shared-staging-transform
  query_params:
    table: regex|customer_[0-9]+
    transform: tenant-transform

Quesma can now be exposed externally to operate with an existing Kibana deployment using the enable_public_access tunable. See Enable public access to Quesma for more information.
Gunicorn will now scale up async workers in accordance with the number of CPUs available to the hdx-traefik-auth container, saving administrators time when adding or removing resources from Traefik. This behavior is overrideable using the hdx_traefik_auth_workers tunable.
Improved how email is handled during cluster creation and password resets, especially for customer-deployed environments.
Cluster creation emails are now only sent to ADMIN_EMAIL (if set). If not set, no email is sent.
Prevented all emails (invites, access codes, password resets) from being sent to CI_USER_EMAIL (e.g. hdx@hydrolix.net or $OWNER@hydrolix.net).
Invite logic now matches password reset logic for email suppression.
Added support for custom SMTP configuration to allow customer-managed delivery.
Added support for updating the patch-level of third party software in internal images. Added patch_date tunable.

Intake and Merge⚓︎

Added support to the Vector software inside the cluster allowing specification of Bearer tokens to accompany messages sent to logs_sink_remote endpoints. This facilitates the use of service accounts to send logs to remote endpoints.
Support for new ip and uuid datatypes has been added. See Support for new ip and uuid datatypes for more information.
Removed the Serialize derivation from objects in the common crate, reducing redundancy in the code.
Added a new kafka_sasl_plain credential type for use with Kafka ingest. The kafka-peer can now authenticate with SASL_SSL using username and password credentials.

Query⚓︎

Replaced the use of min_timestamp and max_timestamp in catalog queries to use a more specific SQL condition. This prevents contradictory conditions from fetching any partitions. (For example, WHERE primary < timestamp AND primary > timestamp should return no results)
Fixed incorrect table name in log messages like the following: db='{db_name}' Storage='TurbineStorage' removed (TurbineStorage being incorrect)
Columns in the GROUP BY clause in summary table SQL may now be removed from the summary table SQL. This allows a high-cardinality column to be removed from a summary table without dropping and re-creating the summary table.
Support for new ip and uuid datatypes has been added. See Support for new ip and uuid datatypes for more information.

Security⚓︎

Set a HYDROLIX_TOKEN cookie using API using the attributes secure, httponly, and samesite. Use /users/current API endpoint to determine expired tokens and clear cookies.
Support for row-level access control. See Notable new features for details.

UI⚓︎

Support for row-level access control. See Notable new features for details.
The UI's browser tab title is now set to the cluster hostname to ease the managment of multiple clusters.
Added a "Column Management" sidebar to the UI in Advanced Options. This feature is only available for non-summary tables.
View and edit current and past column names
Optionally set an alias as the current display name
Add new aliases to existing columns

Bug Fixes⚓︎

Config API⚓︎

The add_names endpoint now adds column names, rather than returning an HTTP 500 error.
Pools-related RBAC permissions are now checked properly so only authorized users can edit resource pools in the API and UI.

Intake and Merge⚓︎

Added the ability to split merge target bounds into smaller chunks. This is an attempt to address an issue with slow merges for late-arriving data in Era 3.
The active field in transform is now optional.
Support for new ip and uuid datatypes has been added.
The new merge controller used to expect strings for the min and max settings for field limits. It now expects integers.
The merge controller now properly handles null or empty shard_key columns in the catalog.
Reattached the duplicate_partitions metric, which was unintentionally disconnected when the AdaptivePartitionSource abstraction was introduced. The metric now tracks duplicate partitions during sourcing and consistently updates in telemetry.
Changed a misleading WARNING message in intake-head to INFO during rejector startup. This reduces noise in logs and prevents confusion during troubleshooting.
No functional changes or documentation impact.
Added a safeguard to prevent a rare buffer overrun panic in the custom CSV reader used by intake-head.
The panic was triggered by an out-of-bounds write during CSV ingestion, likely due to a misaligned buffer size and input row length.
A pre-write check was added to grow the destination buffer as needed.
The patch does not fix the root cause of the excessive buffer size, but avoids the crash and allows the ingestion to recover gracefully.
Also improved support for custom-sized buffer requests by configuring bufio.Scanner to handle larger token lengths as needed.

Cluster Operations⚓︎

hdx-scaler now recognizes its configuration settings.
Fixed the install for hdx-scaler, putting it in /usr/local/bin/ where it's on the system $PATH.
The new cluster login invitation e-mail now give the user the correct URL to log into the cluster.
The ACME init job no longer performs a separate, unneeded DNS resolution check before attempting certificate issuance. This resolves issues where DNS records existed but pods couldn’t see them due to stale cache.
Removed the standalone DNS check from the init-acme container.
Consolidated route verification into the start-lego script using HTTP-based validation.
Mounted the acme-account secret in init-acme to enable reuse of existing account data.
Restructured the ACME logic to be more modular and resilient for both job and cronjob execution.

UI⚓︎

Fixed a bug with disappearing data in the Transform SQL field.
Updated the total volume, compression ratio and compression percent calculations for table details used in UI analysis to correctly use raw_size instead of mem_size.
The Hydrolix UI no longer relies on HYDROLIX_USER_ID and HYDROLIX_ORGANIZATION_ID cookies. Instead, session continuity uses only HYDROLIX_TOKEN,
with user details from the authenticated /current endpoint.
This fixes an issue where the UI would fail to load properly if only the session token cookie was present.
Error messages for bad requests to alter/batch jobs are more descriptive.
Pages that don't use API pagination now show full lists of items.

v5.5.0

Notable new features⚓︎

Row-level access control⚓︎

Column alias management for data tables⚓︎

Support for new ip and uuid datatypes⚓︎

Upgrade instructions⚓︎

Changelog⚓︎

Updates⚓︎

Cluster Operations⚓︎

Intake and Merge⚓︎

UI⚓︎

Improvements⚓︎

Config API⚓︎

Cluster Operations⚓︎

Intake and Merge⚓︎

Query⚓︎

Security⚓︎

UI⚓︎

Bug Fixes⚓︎

Config API⚓︎

Intake and Merge⚓︎

Cluster Operations⚓︎

UI⚓︎

Support for new `ip` and `uuid` datatypes⚓︎