Hi, Are there any circumstances in which a vacuum could corrupt a database?
I am storing about 300000 images in an SQLite database, where each row contains a couple of integers and a BLOB of data (average size 80Kb). There are a some indices on the integer columns, and the final database is around 25Gb on disk. This is a lot of data, but the database is read-only after creation and performance is excellent. Prior to performing a vacuum, the database functions perfectly - selecting data out of a row returns the correct results. Performing a vacuum takes a while, reduces the size of the database, and does not return any errors. After the vacuum, selects trigger sqlite3Corrupt() and sqlite3 returns "SQL error: database disk image is malformed". Running an integrity check immediately after the vacuum shows: ---------------------------------- sqlite> pragma integrity_check; *** in database main *** Page 4198007: sqlite3BtreeInitPage() returns error code 11 On tree page 3597839 cell 61: Child page depth differs Page 4214662: sqlite3BtreeInitPage() returns error code 11 [... 36 similar lines ...] On tree page 4 cell 2: Child page depth differs Page 5082741: sqlite3BtreeInitPage() returns error code 11 On tree page 4 cell 3: Child page depth differs Page 5945673: sqlite3BtreeInitPage() returns error code 11 [... 57 similar lines ...] sqlite> ---------------------------------- I don't think I am near any of the limits mentioned on limits.html, and I am using pragma temp_store_directory to make sure temporary files are going to a disk with several hundred Gb free (monitoring the attached disks shows that only that disk is touched during the vacuum, so I don't think I am running out of space). Normally we vacuum so that we can increase the page size to 32Kb, but omitting that and going straight from a working database with the default page size (1Kb) to a vacuum'd database with the same page size also corrupts the file. The first problem listed is on page 4198007, which with 1Kb pages is a few thousand pages past the 4Gb mark - which may just be a coincidence, as I didn't think there was a 4Gb limit. Is there any reason this operation should fail, like a 4Gb ceiling on some operations? If not I will try and create a test case to reproduce it, but I was wondering if anyone had seen something similar before. -dair ___________________________________________________ d...@refnum.com http://www.refnum.com/ _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
