Changelog
List of aperture PRs merged since 2.18.0 release. For the full list of changes, see [list of changes][changes]
Check if current aperturectl version is newer than on disk (2748)
Description of change
- Add helper functions
- Remove latest and pull latest if current version is newer than on disk
- New Feature: Added version management functionality to `aperturectl`.
The tool now checks if the current version is newer and performs updates
accordingly.
- Refactor: Improved modularity in `blueprints` command. Extracted logic
from `RunE` and `PostRunE` functions into separate, more manageable
functions.
- Refactor: Introduced a new `utils` package for common utility
functions related to version management.
- Chore: Updated descriptions for `RootCmd` and `BlueprintsCmd` commands
for better clarity.
- Bug Fix: Fixed an issue with the removal of blueprints from the local
system in the `blueprints remove` command.
Make sure Tree.Node is always defined (2747)
- New Feature: Introduced `rootTree` variable to hold the result of
`circuitfactory.RootTree()` function, enhancing the structure of the
compiled circuit.
- Refactor: Removed early return condition based on
`policy.evaluationInterval > 0`, simplifying the control flow and making
the function more readable.
- Error Handling: Added `rootTreeErr` to capture any errors returned by
`circuitfactory.RootTree()`, improving error handling and debugging
capabilities.
Remove policy marshalling from aperturectl (2744)
- Refactor: Updated `applyPolicy` and `createAndApplyPolicy` functions
in `cmd/aperturectl/cmd/apply/policy.go` and
`cmd/aperturectl/cmd/cloud/apply/policy.go` to use policy bytes instead
of policy objects, improving flexibility and consistency.
- Refactor: Modified `GetPolicy` and `UpdatePolicyUsingAPI` functions in
`cmd/aperturectl/cmd/utils/policies.go` to handle policy bytes,
enhancing code maintainability.
- New Feature: Changed `UpsertPolicy` function in
`pkg/policies/controlplane/policy-service.go` to support YAML formatted
policies, expanding the range of acceptable input formats.
Handle UNKNOWN service while passing to Check calls (2738)
This change allows both `""` and no `service` field as input in the
selectors and stops reporting `UNKNOWN` as service
- **Refactor**: Simplified YAML comparison in `CompareConfigMap`
function using `MatchYAML` matcher, improving code readability.
- **New Feature**: Added `ParseServicesFromString` and
`ParseServicesFromAddress` methods to the `ServiceGetter` interface,
enhancing modularity and simplifying service retrieval logic.
- **Bug Fix**: Updated `ServicesFromSocketAddress` method to return a
`nil` slice instead of `[]string{"UNKNOWN"}` when IP address is invalid
or no entity is associated with it, providing more accurate results.
- **Refactor**: Replaced direct service retrieval from addresses with
new `ParseServicesFromString` and `ParseServicesFromAddress` methods in
`checkhttp.go` and `authz.go`, improving code maintainability.
- **Chore**: Added logging statements in `checkhttp.go` and `authz.go`
for better visibility of source and destination services.
Preemption metric fixes (2743)
Description of change
* Unregister
* Init the metrics to get zero counters which ensure that the preemption
query which combines all of the metrics returns a valid values even if
some of the counters did not ever increment
- Refactor: Renamed and updated the `Delete` method in the
`SchedulerMetrics` struct for better readability.
- New Feature: Added `appendWorkloadLabel` method to `SchedulerMetrics`
struct, enhancing metric labeling with workload information.
- Refactor: Updated `NewScheduler` function in `Factory` struct to
initialize preempt metrics for each workload, improving metrics
accuracy.
- Refactor: Modified `Decide` method in `Scheduler` struct to use
`getWorkloadLabel` function, providing more precise workload
identification.
Send policy string in the UpsertPolicy req to validate before putting in etcd (2739)
Description of change
This change does not send policy proto in the UpsertPolicy request.
Instead, sends the raw policy string. This way, the string is
unmarshalled to languagev1.Policy on the controller side before putting
in etcd. We get correct validation and proper defaults handling.
- New Feature: Added support for specifying policy as a JSON string in
`UpsertPolicyRequest` with the new `policy_string` field.
- Refactor: Updated package and type names related to policy language
across multiple files for better organization and clarity. The package
`languagev1` has been renamed to `policylangv1`.
- Refactor: Improved handling of policy data in `UpsertPolicy` function,
including better error handling and support for field masks.
- Chore: Replaced usage of `proto.Marshal()` with `MarshalJSON()` for
policy serialization, providing more consistent and efficient data
handling.
add validation checks for deprecated fields (2741)
- **Deprecation**: The `AdaptiveLoadScheduler` component in the
`FlowControl` message has been deprecated. Users are advised to use the
"AIMD Load Scheduler" instead for better performance and efficiency.
- **Deprecation**: The `telemetry_collectors` field in the `Resources`
message and the `differentiator` field in the `Component` message have
been marked as deprecated. Future updates will remove these fields.
- **Documentation**: Updated the configuration specification document to
reflect the deprecation of the "Adaptive Load Scheduler" and
"TelemetryCollector".
- **Refactor**: Updated blueprint metadata to include a new
`deprecation_message` key, providing users with more information about
deprecated components.
use policy name in background circuit job name (2734)
- Refactor: Enhanced the naming convention for background jobs in the
control plane runtime. The job names now include the policy name along
with the policy hash, making them more descriptive and unique. This
change improves traceability and debugging by providing clearer
identification of each job.
Scheduler preemption metrics (2731)
**New Features:**
- Added new metrics related to workload requests and tokens, providing
more detailed insights into system performance. These include
`RequestInQueueDurationMetricName`, `WorkloadPreemptedTokensMetricName`,
`WorkloadDelayedTokensMetricName`, and `WorkloadOnTimeMetricName`.
- Introduced a new `preemptionMetrics` struct in the `WFQScheduler` for
tracking token counters related to preemption and delay metrics.
**Refactor:**
- Updated the `queuedRequest` struct with two new fields:
`tokensInQueue` and `tokensAllowed` to better manage workload requests.
- Enhanced the `queueRequest`, `cancelRequest`, and `scheduleRequest`
functions to handle the new preemption and delay metrics.
**Documentation:**
- Improved clarity and consistency of the metrics generated by Aperture
Agents in the documentation.
aperturectl: apply all command for blueprints (2733)
- New Feature: Introduced `--values-dir` flag to specify a directory
path containing multiple values files.
- New Feature: Modified `--values-file` flag to accept a single file
path instead of a blueprint name.
- New Feature: Added `URIToRawContentURL` function that converts a URI
to a raw content URL, specifically for GitHub.
- Refactor: Renamed `GetPolicyTUIModel` to `GetPoliciesTUIModel` across
the codebase for improved clarity.
- Refactor: Updated logic for handling values files and directories in
`generateCmd` function.
- Refactor: `setupOutputDir` function now creates a temporary directory
if `outputDir` is not provided.
- Documentation: Updated command-line options of the `aperturectl` tool
in the documentation.
Add sections under Adaptive Service Protection based on metrics origin (2577)
Description of change
<img width="1898" alt="postgresql-dashboards-signal"
src="https://github.com/fluxninja/aperture/assets/1553055/cab7cb69-4ca7-42d3-bb26-2c30772a726b">
<img width="1895" alt="postgresql-dashboards-tokens"
src="https://github.com/fluxninja/aperture/assets/1553055/3d7e53de-cbb5-4860-be12-b4c5d60fdff4">
- New Feature: Added new Markdown files for Adaptive Service Protection
use cases, including database, infrastructure, message queue, and
service. These documents provide detailed explanations and guides for
each use case.
- Documentation: Updated various documentation files to improve clarity,
readability, and consistency. This includes changes in wording, fixing
typos, updating links, and rephrasing sentences.
- Refactor: Updated comments and struct field names in
`cmd/aperture-agent/config/types.go` for better clarity.
- Chore: Updated file paths in `operator/hack/create_policy_sample.sh`
and `apertureVersion.js`.
- Documentation: Updated the sidebar position in "Percentage Rollouts"
tutorial page.
- New Feature: Added a new PostgreSQL policy guide in
`docs/content/guides/database-load-management/postgresql.md`.
- Refactor: Modified the `vuStages` array in
`playground/scenarios/graceful-js/load-generator/test.js` to adjust the
target values for different durations.
- Documentation: Updated the load management capabilities section in
`docs/content/introduction.md` with revised feature names and
descriptions.
---------
Co-authored-by: Sudhanshu Prajapati <spsupraj2dgmail.com>
Expose policy hash function (2730)
Fix data race in Controller shutdown (2729)
- Refactor: Simplified the creation of watcher instances in
`policy-factory.go` by removing unnecessary variables and consolidating
the creation process.
- Refactor: Removed the `NotifierLifecycle` function from `watcher.go`,
indicating a refactoring or relocation of the functionality for adding
and removing prefix notifiers.
- Style: Updated job group name from "promql_jobs_scheduler" to
"background_jobs_scheduler" in `background-scheduler.go` for better
clarity and consistency.
change default for delayedInitialFill (2706)
Fix deadlock in peers shutdown (2703)
- Refactor: Improved concurrency in the `PeerDiscovery` struct in
`pkg/peers/peers.go`. Introduced separate locks for `peers` and
`services` to prevent data races and enhance performance.
- Refactor: Renamed methods `RegisterSelf` and `DeregisterSelf` to
`registerSelf` and `deregisterSelf` respectively, enhancing code
readability and maintainability.
- Refactor: Updated various methods in `PeerDiscovery` to use the new
lock fields, improving code consistency and reliability.
reject when req tokens are more than bucketCapacity (2702)
global token bucket: improve wait time precision (2699)
[changes]: https://github.com/fluxninja/aperture/compare/releases/aperture-controller/v2.18.0...releases/aperture-controller/v2.19.0-rc.4