(address . bug-guix@gnu.org)
Hello,
Over the last few weeks I made sure that all Cuirass SQL queries were
using indexes. As the "Builds" and "Outputs" tables can be really large,
having queries covered by indexes is imperative for consistent queries
duration.
However, I observed that some queries have inconsistent duration.
Toggle snippet (10 lines)
2020-10-07T17:59:09 Database worker unresponsive for 5 seconds (db-get-builds-max).
2020-10-07T17:59:13 GET /build/3183151/details
2020-10-07T17:59:13 GET /eval/92
2020-10-07T17:59:14 builds request took 3.66e-4 seconds
2020-10-07T17:59:14 Database worker unresponsive for 5 seconds (db-get-builds-max).
2020-10-07T17:59:15 GET /eval/13234
2020-10-07T17:59:16 GET /build/3146487/details
2020-10-07T17:59:19 Database worker unresponsive for 5 seconds (db-get-builds-max).
"db-get-builds-max" query has the following query plan:
Toggle snippet (6 lines)
QUERY PLAN
|--SEARCH TABLE Builds USING INDEX Builds_stoptime_id (stoptime=?)
`--SCALAR SUBQUERY 1
`--SEARCH TABLE Builds USING INDEX Builds_evaluation_index (evaluation=?)
so it should always reasonably fast. The log seem to indicate that the
worker running one of those queries is busy for more than 10 seconds
here. My understanding of SQLite WAL mode is that writers should not
block readers. So I don't get why this query can be intermittently so
slow.
Thanks,
Mathieu
--