MySQL Bugs: #98530: crash when inplace encryption resumes makes tablespace unusable

Bug #98530	crash when inplace encryption resumes makes tablespace unusable
Submitted:	10 Feb 2020 9:27	Modified:	17 Sep 2020 16:25
Reporter:	Satya Bodapati (OCA)	Email Updates:
Status:	Closed	Impact on me:	None
Category:	MySQL Server: InnoDB storage engine	Severity:	S2 (Serious)
Version:	8.0, 8.0.19	OS:	Any
Assigned to:		CPU Architecture:	Any

Description:
kill server when ALTER TABLESPACE ENCRYPTION='Y'/'N' is in progress.
When recovery happens, kill the server again at the moment it 'resume encryption' operation is about update DD.

Server fails to load tablespace in next startup

2020-02-10T09:20:40.824917Z 7 [ERROR] [MY-012219] [InnoDB] In file 'encrypt_ts.ibd', tablespace id and flags are 4 and 26624, but in the InnoDB data dictionary they are 4 and 18432. Have you moved InnoDB .ibd files around without using the commands DISCARD TABLESPACE and IMPORT TABLESPACE? Please refer to http://dev.mysql.com/doc/refman/8.0/en/innodb-troubleshooting-datadict.html for how to resolve the issue.
2020-02-10T09:20:40.824930Z 7 [ERROR] [MY-012131] [InnoDB] Could not find a valid tablespace file for `encrypt_ts`. Please refer to http://dev.mysql.com/doc/refman/8.0/en/innodb-troubleshooting-datadict.html for how to resolve the issue.

How to repeat:
Apply attached patch. Run attached mtr testcase.

mtr testcase

Attachment: tablespace_encrypt_2_crash.test (application/octet-stream, text), 3.48 KiB.

debug crash point change

Attachment: 98530_code_change.diff (text/x-patch), 539 bytes.

=============================================================================
                  TEST NAME                       RESULT  TIME (ms) COMMENT
------------------------------------------------------------------------------
[100%] innodb.tablespace_encrypt_2_crash         [ fail ]
        Test ended at 2020-02-10 14:50:41

CURRENT_TEST: innodb.tablespace_encrypt_2_crash
mysqltest: At line 77: Query 'SELECT * FROM t1 LIMIT 10' failed.
ERROR 1812 (HY000): Tablespace is missing for table `test`.`t1`.
safe_process[17324]: Child process: 17325, exit: 1

This resume thread exits after it finishes its work. So I couldn't avoid sleeps.. Would ideally avoid sleeps to get deterministic case..
May be a combination of debug_sync and debug_execute_if can avoid that..

Hello Satya,

Thank you for the report and feedback!

regards,
Umesh

Posted by developer:
 
Fixed as of the upcoming 8.0.23 release, and here's the proposed changelog entry from the documentation team:

An interrupted tablespace encryption operation did not update the
encrypt_type table option information in the data dictionary when the
operation resume processing after the server was restarted.