After upgrading a PostgreSQL server to SLES 15 SP3, a REINDEX is required.
This document (000020305) is provided subject to the disclaimer at the end of this document.
Before using the database productively again after the upgrade, text indices must be re-indexed, or there may be corruption on the first write access to the database.
Reading from the DB does not do any harm to the stored data, but the results may be unexpected or even wrong.
Note: This can take considerable time depending on size and number of the indices!
Reindexing can be done via "reindexdb" wrapper packaged with postgresql.
To reindex the table foo in a database named bar:
$ reindexdb --table=foo bar
Reindex all tables in a database:
$ reindexdb -a
$ man reindexdb $ man 7 reindex
All subsequent Service Packs and Products will require the same precaution if the DB had been created on a system with a glibc older than 2.28!
Actually there is no bug, but it is a change in behavior which can cause a great deal of trouble.
C” or “
POSIX” collations should not be affected.
It does not matter which PostgreSQL version is in use, as all versions can be affected.
This collation issue also affects physical (WAL) replication which is supported by the PostgreSQL core itself, where all systems in a replicated cluster must have the same collations in place!
Since version 10, PostgreSQL does not only support “libc collations” but also “ICU collations” which are not affected by this particular problem. See:
SUSE Manager Users: Please read:
- Document ID:000020305
- Creation Date: 28-Jun-2021
- Modified Date:29-Jun-2021
- SUSE Linux Enterprise Server
For questions or concerns with the SUSE Knowledgebase please contact: email@example.com