All new instances uses the utf8mb4 charset. Existing databases are not migrated as the operation is slow and can potentially be destructive. For this reason, we keep the existing tests making sure the table can be still be created with both the utf8 and utf8mb4 charset (it could also help detecting edge cases regarding the management of the charset between DB engines). It is mandatory that the DB server have the innodb_large_prefix option enabled. This option is the default in MySQL 5.7 and is the only possible choice in MySQL 8.0. Out of precaution, Mediawiki databases are for now left untouched. It is possible to manually convert an existing database with something like this (see [0]): ``` $ /opt/rh/rh-mysql57/root/usr/bin/mysqldump --skip-set-charset --no-data --databases <tuleap_db> > schema.sql $ /opt/rh/rh-mysql57/root/usr/bin/mysqldump --skip-set-charset -n -t --databases <tuleap_db> > data.sql $ sed 's/DEFAULT CHARACTER SET latin1/DEFAULT CHARACTER SET utf8mb4/g' schema.sql > schema_utf8mb4.sql $ sed -i 's/DEFAULT CHARSET=utf8/DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_unicode_ci /' schema_utf8mb4.sql $ sed -i 's/SET character_set_client = utf8/SET character_set_client = utf8mb4/' schema_utf8mb4.sql $ /opt/rh/rh-mysql57/root/usr/bin/mysql < schema_utf8mb4.sql $ /opt/rh/rh-mysql57/root/usr/bin/mysql < data.sql ``` [0] https://www.percona.com/blog/2018/04/10/migrating-database-charsets-to-utf8mb4/ Change-Id: I9b139ef79b3da904198fe2a6d99c064e75d99b12

+75 −4

Author

Thomas Gerbet (tgerbet) 2020-06-26 13:53

Committer

Thomas Gerbet (tgerbet) 2020-06-26 13:53

Hash a579455b06f4983df50cfe41665c0aac1041c459

Parent 331b4a3e1f1df45506db387eab48cd259ffd1b30

Reference git #tuleap/stable/a579455b06f4983df50cfe41665c0aac1041c459