database: Use "write-ahead log" mode and set a long "busy timeout".
This should avoid "database is locked" errors when there's a lot of concurrency, for instance when offloading simultaneously a lot of builds. * guix/store/database.scm (call-with-database): Add two 'sqlite-exec' calls to set 'journal_mode' and 'busy_timeout'.
This commit is contained in:
parent
b96e05aefd
commit
bdf860c2e9
|
@ -79,6 +79,15 @@ as specified by SQL-SCHEMA."
|
|||
create it and initialize it as a new database."
|
||||
(let ((new? (not (file-exists? file)))
|
||||
(db (sqlite-open file)))
|
||||
;; Turn DB in "write-ahead log" mode, which should avoid SQLITE_LOCKED
|
||||
;; errors when we have several readers: <https://www.sqlite.org/wal.html>.
|
||||
(sqlite-exec db "PRAGMA journal_mode=WAL;")
|
||||
|
||||
;; Install a busy handler such that, when the database is locked, sqlite
|
||||
;; retries until 30 seconds have passed, at which point it gives up and
|
||||
;; throws SQLITE_BUSY.
|
||||
(sqlite-exec db "PRAGMA busy_timeout = 30000;")
|
||||
|
||||
(dynamic-wind noop
|
||||
(lambda ()
|
||||
(when new?
|
||||
|
|
Loading…
Reference in New Issue