github.com/VictoriaMetrics/VictoriaLogs

Released at 2026-10-02

v1.52.x is a line of LTS releases. It contains important up-to-date bugfixes for VictoriaLogs enterprise.
All these fixes are also included in the latest community release.
The v1.52.x line will be supported for at least 12 months since v1.52.0 release

• SECURITY: upgrade Go builder from Go1.26.5 to Go1.26.8. See the list of issues addressed in Go1.26.6, Go1.26.7 and Go1.26.8.

• SECURITY: deletion API: restrict the /delete/run_task endpoint to the POST method only in order to prevent some SSRF-based log deletion attacks. See #1635.

• BUGFIX: Loki data ingestion: properly ignore structured metadata with empty values in protobuf requests. Previously, valid requests containing structured metadata with empty values were rejected. See #1757.

• BUGFIX: metrics: prevent vl_streams_created_total from decreasing when an old daily partition is removed. The metric now increases until restart, so Prometheus increase() no longer reports false spikes after old partitions are deleted. See #1461.

• BUGFIX: syslog data ingestion and unpack_syslog pipe: prevent a panic when an incomplete RFC5424 structured data parameter ends immediately after =. This could occur both when ingesting syslog messages and when parsing already stored messages with unpack_syslog. See #1786.

• BUGFIX: syslog data ingestion: rename the misspelled vl_udp_reqests_total metric to vl_udp_requests_total, so it matches the docs.

• BUGFIX: cluster version (vlinsert): now drains buffered logs to vlstorage nodes on graceful shutdown instead of dropping them, bounded by the new -insert.drainTimeout command-line flag (default 5s). See #1572.

• BUGFIX: cluster version: evenly spread rerouted data across available vlstorage nodes. Previously, healthy nodes adjacent to unavailable nodes in the -storageNode list could receive much more data, resulting in uneven resource usage. See #1548.

• BUGFIX: cluster version: properly cancel queries already running on vlstorage when the corresponding query is canceled or times out on vlselect. Previously, vlstorage could fail to detect the disconnected vlselect, causing these queries to continue running and waste CPU and disk IO. See #1672.

• BUGFIX: data ingestion and querying: properly handle logs containing duplicate stream field names. Previously, v1.52.0 could panic when ingesting such logs in single-node VictoriaLogs, drop them during ingestion in VictoriaLogs cluster, or panic when querying such data written by earlier releases. See #1603 and #1604.

• BUGFIX: LogsQL: fix week_range[Sun,Sun] filter when it is used inside the filter pipe. Previously, it could fail to match rows on Sunday. See #1335.

• BUGFIX: LogsQL: properly execute queries returning the last N logs (such as queries ending with | sort by (_time desc) limit N) when they contain pipes writing to the _time field, such as math ... as _time, replace (...) at _time or extract "<_time>". Previously such queries could fail with the missing _time field in the query results error or return logs in the wrong order when they were executed on wide time ranges. See #1727.

• BUGFIX: LogsQL: fix sort by (_time) limit N returning logs out of order when the query pipeline included pipes like unpack_json that overwrite _time. See #1360.

• BUGFIX: LogsQL: fix a crash when the math pipe contains a quoted constant such as "2025-01-01T00:00:00Z", and the query is executed with the limit query arg (the web UI always sets it) or via live tailing.

• BUGFIX: web UI: render only inline Markdown links with explicit http or https destinations, such as [text](https://example.com), as clickable links in log messages. Bare URLs, autolinks, reference-style links, links using other schemes, and images are rendered as plain text. See #1470.

• BUGFIX: File Collector: ignore permission denied error when trying to find a rotated log file from previous runs during startup. Previously, vlagent failed with non-zero error code if it couldn't open a file in the same directory as log file. See #1796.

• BUGFIX: File Collector: properly set the file field for logs collected from multiple files matching the same -fileCollector.glob when -fileCollector.extraFields is set. Previously, depending on the number of extra fields, logs from one file could be stored with the file field and the log stream of another file. See #1818.

• BUGFIX: File Collector: incomplete last log lines are no longer joined with the first line of the rotated file. They are dropped with a warning instead. This matches the behavior of other log collectors. See #1819.

• BUGFIX: File Collector: drop the incomplete last log line with a warning when the log file is deleted. Previously, vlagent treated this case as a bug and panicked. See #1552.

Released at 2026-08-18

• SECURITY: upgrade Go builder from Go1.26.4 to Go1.26.6. See the list of issues addressed in Go1.26.5 and the list of issues addressed in Go1.26.6.

• BUGFIX: cluster version: allow vlstorage v1.51.1 to accept requests from vlselect versions v1.38.0 through v1.51.0. This allows upgrading vlstorage nodes from older versions to v1.51.1 while vlselect nodes remain on the corresponding older version, and then upgrading the vlselect nodes to v1.51.1, without query downtime caused by an internal protocol mismatch. Starting from v1.52.0, vlstorage no longer accepts requests from vlselect v1.50.0 or earlier. See #1665.

Released at 2025-12-26

• BUGFIX: /select/logsql/facets endpoint: fix expecting 3 columns; got 4 columns panic when /select/logsql/facets endpoint is queried at cluster version of VictoriaLogs. See #940. The issue has been introduced in this commit in the release v1.35.0.
• BUGFIX: Kubernetes Collector: properly handle "410 Gone" responses from the Kubernetes API server. Previously, vlagent could report 410 errors and fail to collect logs from newly discovered Pods. The bug has been introduced in v1.43.0.
• BUGFIX: HTTP querying API: properly add labels from the by (...) clause to responses returned from /select/logsql/stats_query and /select/logsql/stats_query_range endpoints for histogram() stats function results. See this comment for details.

Released at 2025-12-22

• FEATURE: Kubernetes Collector: add an ability to include and exclude Pod/Node labels and annotation fields from logs. This metadata is also available for filtering. See #923 and these docs for details.

• FEATURE: Kubernetes Collector: reduce a load on the Kubernetes API by using resource versions to ignore already processed events when a TCP connection is dropped. In the previous version, vlagent would request the full state of the current node from the control plane every 5 minutes, which could lead to an increased load on the central API server in large clusters.

• BUGFIX: vlstorage: fix incorrect warning logs about time field in /internal/insert and /internal/native APIs. Thanks to @cyberzz-dev for the issue #937.

Released at 2025-12-20

• FEATURE: HTTP querying API: convert histogram stats results to VictoriaMetrics histogram buckets in responses from /select/logsql/stats_query and /select/logsql/stats_query_range endpoints. This allows using histogram() stats in recording rules for VictoriaLogs. See #10184 at VictoriaMetrics, #86 and #62 at VictoriaLogs plugin for Grafana.

• FEATURE: Kubernetes Collector: add support for metadata-based filtering for Kubernetes logs via -kubernetesCollector.excludeFilter command-line flag. The flag accepts any LogsQL filter which is applied to Pod labels, namespaces, and other metadata before reading the log files. This enables early exclusion of unwanted log records, reducing CPU usage and I/O overhead. See #923 and these docs for details.

• FEATURE: Kubernetes Collector: automatically parse Kubernetes system logs as structured data. Previously, vlagent wrote these logs as unstructured text.

• FEATURE: vlagent: add -tmpDataPath command-line flag, that defines default root path for all vlagent data, that requires persistence: remote write persistent queue, collector checkpoints path.

• FEATURE: vlinsert: implement /insert/native endpoint for public access to the internal VictoriaLogs ingestion protocol. See #38 and these docs.

• FEATURE: web UI: add localStorage availability checks with error reporting. See #843.

• FEATURE: web UI: add VLUI:-prefixed localStorage keys and legacy key migration.

• BUGFIX: web UI: fix incorrect y-axis scaling in stats charts when displaying negative values. See #905.

Released at 2025-12-17

• BUGFIX: data ingestion: accept timestamps with timezone offsets in ±HHMM format (e.g. 2025-12-15T02:12:34.977+0100) in addition to ±HH:MM (RFC3339) format. See #902.
• BUGFIX: OpenTelemetry data ingestion: fix a memory leak during ingestion of logs with complex attributes. See this comment at #869.

Released at 2025-12-13

• FEATURE: LogsQL: return _stream field from block_stats pipe. This allows detecting log streams, which occupy the most of disk space according to this query.

• FEATURE: OpenTelemetry data ingestion: parse scope inside ScopeLogs and store it into log fields starting with scope. prefix. See #826.

• FEATURE: ElasticSearch data ingestion: return stub responses for the requests to /insert/elasticsearch/_rollup* endpoints. Thanks to @Sakalya for the pull request #890.

• FEATURE: web UI: implement "Stats view" mode for the graph using stats_query_range to display log stats. See #34.

• BUGFIX: OpenTelemetry data ingestion: properly handle null values inside arrays. Previously, such arrays could cause a panic during ingestion. See #869. The issue has been introduced in v1.40.0.

• BUGFIX: OpenTelemetry data ingestion: properly skip attributes with missing key. See this report.

• BUGFIX: OpenTelemetry data ingestion: reduce memory usage after v1.40.0 release. See #869.

• BUGFIX: Kubernetes Collector: properly handle unset -kubernetesCollector.msgField and -kubernetesCollector.timeField flags. Previously, vlagent didn't apply the default values described in the documentation.