[MDEV-16203] autoinc_debug of rocksdb test suite fails Created: 2018-05-17  Updated: 2018-08-12  Resolved: 2018-08-12

Status: Closed
Project: MariaDB Server
Component/s: Storage Engine - RocksDB
Affects Version/s: 10.2, 10.3
Fix Version/s: 10.2.17, 10.3.9

Type: Bug Priority: Major
Reporter: Oleksandr Byelkin Assignee: Sergei Petrunia
Resolution: Fixed Votes: 0
Labels: None


 Description   

CURRENT_TEST: rocksdb.autoinc_debug
--- /home/sanja/maria/git/server/storage/rocksdb/mysql-test/rocksdb/r/autoinc_debug.result	2018-05-17 11:53:27.913983298 +0200
+++ /home/sanja/maria/git/server/storage/rocksdb/mysql-test/rocksdb/r/autoinc_debug.reject	2018-05-17 13:13:59.953187229 +0200
@@ -61,10 +61,10 @@
 ERROR HY000: Lost connection to MySQL server during query
 select table_schema, table_name, auto_increment from information_schema.tables where table_name = 't';
 table_schema	table_name	auto_increment
-test	t	4
+test	t	1
 select max(i) from t;
 max(i)
-3
+NULL
 # After engine prepare
 begin;
 insert into t values ();
@@ -74,10 +74,10 @@
 ERROR HY000: Lost connection to MySQL server during query
 select table_schema, table_name, auto_increment from information_schema.tables where table_name = 't';
 table_schema	table_name	auto_increment
-test	t	4
+test	t	1
 select max(i) from t;
 max(i)
-3
+NULL
 # After binlog
 begin;
 insert into t values ();
@@ -87,10 +87,10 @@
 ERROR HY000: Lost connection to MySQL server during query
 select table_schema, table_name, auto_increment from information_schema.tables where table_name = 't';
 table_schema	table_name	auto_increment
-test	t	6
+test	t	3
 select max(i) from t;
 max(i)
-5
+2
 # After everything
 begin;
 insert into t values ();
@@ -100,8 +100,8 @@
 ERROR HY000: Lost connection to MySQL server during query
 select table_schema, table_name, auto_increment from information_schema.tables where table_name = 't';
 table_schema	table_name	auto_increment
-test	t	8
+test	t	5
 select max(i) from t;
 max(i)
-7
+4
 drop table t;
 
mysqltest: Result length mismatch



 Comments   
Comment by Sergei Golubchik [ 2018-05-17 ]

Doesn't fail for me when running standalone. Fails if I run ./mtr --parallel 5 --suite rocksdb

Comment by Sergei Petrunia [ 2018-08-12 ]

The code for crash_commit_before:

  DBUG_EXECUTE_IF("crash_commit_before", DBUG_SUICIDE(););

The state of the database after restart seems to be as if the last few transactions before the "crash_commit_before" was hit were not persisted on
disk (playing around with the testcase confirms this).

MariaDB runs the rocksdb testsuite with rocksdb-flush-log-at-trx-commit=0. This is a difference from the upstream. The reason for this was that this is irrelevant for most of the tests, and without this a lot of tests take unacceptably long on buildbot/other dev machines).

Generated at Thu Feb 08 08:27:08 UTC 2024 using Jira 8.20.16#820016-sha1:9d11dbea5f4be3d4cc21f03a88dd11d8c8687422.