Persistent ‘search-max-unique-timeseries’ via TiUP #2682
Description
Background:
https://tidb.atlassian.net/browse/FRM-3490?focusedCommentId=20519295
Context-
TopSQL Error:
[2026/01/10 09:19:55.336 +08:00] [WARN] [default_query.go:355] ["failed to fetch timeseries db"] [error="{"status":"error","errorType":"422","error":"search error after reading 0 data blocks: error when searching for tagFilters=[{name=\"cpu_time\",instance=\"10.107.41.5:10080\",instance_type=\"tidb\"}] on the time range [2026-01-09T15:59:55Z..2026-01-09T16:16:00Z]: error when searching for metricIDs in the current indexdb: the number of matching timeseries exceeds 300000; either narrow down the search or increase -search.max* command-line flag values (the most likely limit is -search.maxUniqueTimeseries); see https://docs.victoriametrics.com/#resource-usage-limits"}"]
This can be mitigated by manually adjusting the parameter:
[tsdb]
search-max-unique-timeseries = 6000000
However, these changes are currently not persistent and are overwritten on restart or upgrade. As a result, it forces users to repeatedly reapply the same modification.
Feature Request
Please allow search-max-unique-timeseries to be configured via cluster configuration, so that:
Users can adjust this limit according to their cluster size and workload characteristics
The configuration is persisted and automatically applied across restarts and upgrades