Release notes for August 2026

Grepr released these feature updates, improvements, and bug fixes in August 2026.

August 26, 2026

Mask sensitive values on any log pipeline, and only where you need to

You can now add a masking step to any log pipeline rather than only to pipelines built from a template, and give each mask an optional condition so that only the logs matching it are rewritten. A conditional mask leaves every other log untouched, which lets you redact a value that appears in one service’s logs without rewriting the rest of the pipeline’s traffic. Masking runs before logs reach the Grepr data lake and your vendor sinks, so the masked value is both what Grepr stores and what it forwards.

A reorganized log reducer editor

The Edit Reducer dialog is now three focused sections instead of one long form. Basic holds the reduction window, similarity threshold, and minimum samples. Fields holds one searchable table where you set grouping and merge strategy per field. Advanced holds masks, delimiters, sampling, and output switches. Each section opens on its own and fits on a single screen, so the settings you change most often are the ones you reach first. See Configure the log reducer in a Grepr pipeline.

Other improvements and fixes

• Log Explorer now applies instant filtering to queries that filter on status or level, and a query it cannot filter in the browser reports that plainly instead of showing an internal translator error.
• Improved the performance of trace pipelines.
• Fixed an issue where a backfill to a dataset or vendor integration created before Grepr adopted its current identifier format was rejected as invalid, so the backfill never started.
• Fixed an issue where the actions menu on a field in the Log Explorer detail panel opened off the edge of the screen, and where choosing an item from it closed the detail panel.

August 20, 2026

Grok rules that name an unrecognized transformer are now rejected

Grepr used to skip a transformer it did not recognize and parse the rest of the rule, so a misspelled transformer silently left the extracted value as a raw string. A rule that names an unrecognized transformer is now reported as an error, both when you test the rule and when you start the pipeline holding it. Testing a rule also reports an error for a regular expression construct Grepr cannot match, such as an atomic group (?>...). See The Grepr Grok parser: Supported transformers.

Tune how a dataset stores its raw logs

You can now edit the partitioning and the promoted tag columns of a dataset’s raw log table from the dataset’s settings. Adding hashed bucket partitions on the fields you filter by most, and promoting tag keys to top-level columns, speeds up the searches and backfills you run against that dataset. This was previously limited to selected organizations and is now available to every organization.

Other improvements and fixes

• Grepr now splits log events larger than 900KB, reduced from 1MB.
• The pipelines, jobs, datasets, and log search pages load faster and repeat far fewer requests, and a browser tab in the background no longer keeps refreshing pipeline charts and job lists.
• Fixed an issue where, in rare circumstances, a pipeline assigned to one or more teams became assigned to the organization instead.
• Fixed an issue where tail sampling on a trace pipeline dropped the shortest trace recorded for a trace signature instead of keeping it at the lowest configured tier, which made the set of traces kept depend on the order they arrived in.

August 15, 2026

Query and backfill spans with the Grepr CLI

You can now run grepr query and grepr backfill against span data as well as logs by adding --data-type spans, or by passing a trace pipeline to --job-id, which infers the data type from the pipeline. A span backfill replays raw spans from the Grepr data lake to your Datadog or OpenTelemetry trace destinations, selected with the same Datadog-like syntax you use in Trace Explorer, and Grepr tracks which spans it has already delivered so that repeating a backfill does not send them twice. See The Grepr CLI.

Other improvements and fixes

• The Datadog integration setup guide now includes examples for the Datadog Agent deployed with the Datadog Helm chart and with the Datadog Operator, and every collector example shows how to dual ship by adding Grepr as an additional endpoint while your existing Datadog intake stays in place. See Configure Datadog connections with a Grepr integration.
• Spans now carry an identifier for their trace signature, which Trace Explorer uses to group spans by signature. Spans stored before this release do not have this identifier, so signature grouping and the tree view are unavailable for them.
• Backfilled logs now carry the grepr.backfilled and grepr.backfilled.timestamp tags as event tags rather than as vendor sink tags, and the timestamp value is the end of the backfill time range instead of the time you created the backfill.
• Reduced the storage that the Grepr data lake uses to track which spans a backfill has already delivered.
• The text field that confirms a deletion now has a label, so assistive technologies can announce what it is for.
• Fixed an issue where auto-synced exceptions were deleted when Grepr could not read some of the monitors or dashboards in your observability vendor account, such as when an access key lacked permission for a dashboard or the vendor rate limited the request. See Prevent aggregation of messages used in your observability platform.
• Fixed several issues where searching logs by tag matched the wrong records, including a tag holding more than one value failing to match an exact search, wildcard and prefix searches missing records, a negated search such as -env:prod excluding records that carry no env tag at all, and tag values containing a comma being split into separate values.
• Fixed an issue where two queries translated at the same time could interfere with each other, so a search could lose part of its filter and return records it should have excluded, or fail while being planned.
• Fixed an issue where Log Explorer results stopped loading when a log contained a number of 22 digits or more.
• Fixed an issue where filtering on an attribute whose value has mixed-case letters, such as @env:Prod, missed matching records in Log Explorer.
• Fixed an issue where a backfill that overlapped one already running was allowed to start, which could deliver the same logs to your vendor twice, and extended the same protection to span backfills.
• grepr_pipeline_bytes_out_per_minute now reports the bytes Grepr delivered to your observability vendor sinks in the last complete minute instead of always reporting 0, and a pipeline that writes only to the Grepr data lake reports no series for this metric. See Monitor performance metrics.
• Fixed an issue where the teams you selected on a Datadog integration were discarded when you saved, and where an unsaved team selection could appear on the integration card as though you had saved it.
• Fixed an issue where a pipeline that combines several inputs could leave some of its processing tasks without work, which reduced throughput.
• Fixed an issue where updating a stopped pipeline reported the update as failed even though it succeeded.
• Fixed inconsistent text sizes, weights, and line spacing across the Grepr UI, including a delete confirmation dialog whose message rendered smaller than the text around it.

August 4, 2026

Cut span volume with drop and merge rules on trace pipelines

You can now add span reduction rules to a trace pipeline to reduce what you send to your vendor while the complete trace stays queryable in the Grepr data lake. A drop rule removes spans you do not need along with everything beneath them, and a merge rule collapses repeated identical sibling spans, such as those produced by a fan-out or a loop, into one representative span carrying a count. Drop and merge rules are configured independently. See Use a trace reduction pipeline for your APM data.

Add multiple API keys to a Datadog integration

A Datadog integration can now hold more than one API key, so an existing fleet of Datadog keys can route through a single Grepr integration instead of needing one integration per key. Any key on the integration authenticates incoming telemetry, keys carry optional labels and rotate independently, and one designated send key is used for the data Grepr writes back to Datadog. See Configure Datadog connections with a Grepr integration.

Other improvements and fixes

• A query that references a template variable, such as host:__host, now matches instead of being read as literal text, and a negated clause such as -sometag:value now also matches records where the tag is absent, following the Datadog behavior.
• Fixed an issue where a span without a service name could stop the pipeline reading it, and a malformed JSON record caused the rest of the file to be skipped without being reported.
• Fixed an issue where an hour of usage could be lost or recorded against the wrong hour, which affected the daily usage table, your commitment figures, and your invoice.
• Fixed several issues in the Grepr CLI, including streaming CSV export misaligning columns and cutting output short, job:get ignoring --resolved, saved timezone and authentication settings being overwritten by defaults, and generated Splunk backfill links covering slightly the wrong time range.
• Fixed an issue where several endpoints rejected a request that omitted an optional parameter, which affected paging through vendor exceptions, searching activity logs, SSO claim mappings, and role updates.
• Fixed an issue where impact estimation could fail on a high-volume dataset, which left the passthrough estimate for that dataset out of date.
• Fixed an issue where a pipeline that could not resume from its saved state had that state deleted, which removed the snapshot you would have rolled back to.
• Fixed an issue where each update to a running pipeline left connections open, so a pipeline updated many times ran out of memory and restarted on its own.
• Blank sessionization durations and blank span sampling rates are now rejected when you submit them instead of being saved as invalid values.
• Fixed an issue where time range presets in search resolved to the wrong window, so a range of yesterday produced a rolling 24-hour window rather than the previous calendar day.