Argus Versions Save

Argus 2.21.3 release is in production. Here are the features/changes released

Ajna consumer changes to eliminate duplicate values with the same timestamp. Metric meta data indexing improvements to identify high cardinality metrics Improvements to metric discovery query time Enforce max datapoints limit on fill transform output Bug fix NO_DATA trigger not firing when expression is wrapped in SUM transform Templatize notification fields of alerts Bug fix . Alert notification message at times is saying alert is triggered even though alert is cleared Metric name length restriction to 255 character.

• Add metrics for bloom filter
• Use SchemaCommitter hostname hash so that a separate config is not needed
• Use create to add metrics for scopenames and scopemetricnames and update if create fails. This will populate cts field
• Argus downsample function is now accepts extra parameters for start/end times and default value. These values are used to fill gaps in the downsampled period whenever data is missing
• Added a link to actual metric expression url corresponding to the absolute time when the alert was triggered in the alert notification
• Added ability to customize alert name, notification name and custom text content based with variables like ${scope}, ${metric} and ${device} which will get substituted at runtime when the alert gets triggered

• Argus internal KPI for time to detect the alert end-to-end
• Argus sends notifications about missing data to notification owner only
• Formatting changes for Alert notification
• Supporting customized argus alert trigger name
• New paginated APIs to fetch user/shared/private alerts (part1: new APIs, no caller yet).
• Argus WS returns 500 instead of 400 for an invalid metrics expression

1.Add Oauth Support for Authentication 2.Customized argus alert trigger name 3. Formatting changes for Alert notification

1. Discovery Lookup for Scope in separate index (aka faster discovery lookups for scope)
2. BugFix: Metric reducer transform to not throw ArrayOutOfBoundsException when operating on empty timeseries

Alert scheduling improvements - a. Reduce the alert scheduling interval from 15 minutes to 1 minute b. Cache alert definitions during scheduling for better performance c. Schedule by grouping cron expressions and predetermining which alerts need to be run at a particular minute, instead of relying on quartz scheduler. d. Add more KPIs to better track the alert scheduler performance

Alert evaluation improvements - a. Caching notification cooldown and trigger status locally on alert clients to prevent too many queries being made to postgres b. Created a whitelist of scopes for which the alerts will not be disabled during a data lag

Add metrics to track usage of discovery service and TSDB service by user

• No. of discovery queries made by user
• No. of expanded discovery results by user
• No. of TSDB Time serieres returned by user

Track Discovery queries that is above a configurable static threshold

Make bloom filter parameters configurable Flush bloom filter at fixed time of day based on configurable property

Minor bug fixes and optimizations

1.Bug fix: unable to render alert list after edit an alert 2.Allow UI to read big numbers in current locale 3.Fix conflicting logger implementations error 4.Add modified timestamp field while indexing in Elastic Search to allow purging of metric schema 5.Addressed third party library vulnerabilities. 6.Add a configurable TSDB Read ConnectionReuseStrategy 7.Add new metrics for quota / debugging 8.Use BloomFiliter for caching in schema clients replacing the existing trie

1. Add a feature to disable alert evaluations when a data lag is detected in the metrics pipeline
2. Add provision to be able to store audit information in HBase instead of postgres
3. Publish the alert evaluation latency as a metric for each alert when the evaluation succeeds
4. Bug fix to reset the Notification cooldown period when an alert is cleared
5. New transform GROUPBYTAG which groups different metrics together using the given tag values and then executes the provided Transform on the grouped set of metrics
6. Add time window tag for user query in read latency KPI metric
7. Add number of expanded time series for user query in read latency KPI metric

• Fix bugs introduced due to serdes of entire alert object
• Optimize retrieval of notification fields (by using native sql instead of JPQL which performs humungous joins) during alert evaluation
• Improvements to dashboard and alert APIs returning list of objects
• TSDBService federation implementation
• Increase no. of allowed tags for metrics from 7 to 50
• Other bug fixes and improvements

• Heatmap visualization type
• Replace Apache Httpclient with ES Restclient in ElasticSearchSchemaService
• ag-table enhancements
• ag-date local timezone support
• ag-flag support in d3.js version
• New annotation label
• Bug Fixes