Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.5, 10.6, 10.11, 11.1(EOL), 11.2, 10.4(EOL), 11.3(EOL)
-
None
Description
InnoDB fails to find the space id to restore the page from doublewrite
buffer. It is not easy to find the space id from consecutive pages when
we don't have page size information (fsp_flags also 0)
Patch to repeat the failure:
diff --git a/mysql-test/suite/innodb/t/doublewrite.test b/mysql-test/suite/innodb/t/doublewrite.test
|
index 89e6577b434..7eedfc7e636 100644
|
--- a/mysql-test/suite/innodb/t/doublewrite.test
|
+++ b/mysql-test/suite/innodb/t/doublewrite.test
|
@@ -65,8 +65,9 @@ my $page_size = $ENV{INNODB_PAGE_SIZE};
|
my $page;
|
do "$ENV{MTR_SUITE_DIR}/../innodb/include/crc32.pl";
|
open(FILE, "+<", $fname) or die;
|
-sysseek(FILE, ($page_size/2), 0);
|
-syswrite(FILE, chr(0) x ($page_size/2));
|
+#sysseek(FILE, ($page_size/2), 0);
|
+#syswrite(FILE, chr(0) x ($page_size/2));
|
+syswrite(FILE, chr(0) x $page_size);
|
sysseek(FILE, 3*$page_size, 0);
|
sysread(FILE, $page, $page_size)==$page_size||die "Unable to read $name\n";
|
sysseek(FILE, 3*$page_size, 0)||die "Unable to seek $fname\n";
|
error message in log file:
2023-12-07 22:32:59 0 [Note] InnoDB: Starting crash recovery from checkpoint LSN=49621,49621
|
2023-12-07 22:32:59 0 [Note] InnoDB: Header page consists of zero bytes in datafile: ./test/t1.ibd, Space ID:0, Flags: 0
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size:1024. Pages to analyze:64
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size: 1024. Possible space_id count:0
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size:2048. Pages to analyze:64
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size: 2048. Possible space_id count:0
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size:4096. Pages to analyze:40
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size: 4096. Possible space_id count:0
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size:8192. Pages to analyze:20
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size: 8192. Possible space_id count:0
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size:16384. Pages to analyze:10
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size: 16384. Possible space_id count:0
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size:32768. Pages to analyze:5
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size: 32768. Possible space_id count:0
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size:65536. Pages to analyze:2
|
2023-12-07 22:32:59 0 [Note] InnoDB: Page size: 65536. Possible space_id count:0
|
2023-12-07 22:32:59 0 [ERROR] InnoDB: Datafile './test/t1.ibd' is corrupted. Cannot determine the space ID from the first 64 pages.
|
2023-12-07 22:32:59 0 [ERROR] InnoDB: Tablespace 5 was not found at ./test/t1.ibd.
|
2023-12-07 22:32:59 0 [ERROR] InnoDB: Set innodb_force_recovery=1 to ignore this and to permanently lose all changes to the tablespace.
|
Solution:
Find the 0th pages in doublewrite buffer
Extract the space id and flags from page0
Use flags to find page_size information and read consecutive pages to match space id
If all pages are matching with doublewrite buffer page0 then write into the datafile.
Attachments
Issue Links
- relates to
-
-
- Closed
-