The following bug has been logged online:
Bug reference: 5602
Logged by: Chris
Email address: christoph.springer@magicinternet.de
PostgreSQL version: 9.0beta3
Operating system: Debian 5.0.3
Description: Recovering from Hot-Standby file backup leads to the
currupted indexes
Details:
A Hot-Standby database is being backed up to a new machine like this:
0. pause WAL file deletion from Hot-Standby
1. backup pg_control
2. sync data and tablespace files
3. wait for checkout using pg_controldata and waiting for UNDO address
change
4. copy all WAL files
5. recover the newly created database using WAL files
The database recovers and starts normally, but some of the indexes are
corrupt afterwards. I could not find any strange log record in logs for the
recovered database.
This is a definition of a index that is corrupt afterwards:
CREATE INDEX idx_tbl_textcol1_textcol2
ON tbl
USING btree
(textcol1, textcol2)
TABLESPACE data5_tbs;
Thank you in advance,
Chris
