v1.3

New features and improvements

• Added outputs with index sizes before and after the rebuild, and separate summarizing outputs after finishing each database and whole maintenance (7e50003, c8bb00a)
• Reduce the lock_timeout to 1 second (3d371a7)
• Added Retries (#5, #7, 9faf4b4)
  • Now every index will retry to reindex after failure (3 attempts), for example, when canceling a statement due to lock timeout.
  • If all attempts failed - the index is considered failed to reindex. The script will not stop with an error and will continue working with the next index.
  • New parameter --failed_reindex_limit. The maximum number of reindex errors during database maintenance (default: 1)
• Code refactoring (5f18d94)

Fixes

• Slightly reworked database connection parameters in first checks to avoid some issues when connecting to remote Postgres (such as located in clouds) (0bc336c, 2caea5b)
• Fixed a bug which raised when using repack on the index with uppercase letters in its name (12f19cf)
• Changed pg_repack error level to avoid skipping timeout errors (d34eded)

Tests:

• add ShellCheck (96e65e4, 53c7a2b)

Thanks, @crztssr for contributing!

Comparing changes: v1.2...v1.3

v1.2

• drop INVALID index with suffix _ccnew (if exists) after an error performing REINDEX INDEX CONCURRENTLY (feebc35, 05ca13e, 80bb6bb, 4d3bbba)
• check for invalid temporary indexes with the suffix "index_" after performing pg_repack (0a5888e, 0429266)
• check PostgreSQL availability before maintenance each index (a4c9195)
• do not continue working if an error has occurred (1138e5b)
• add the "-X" option to the psql command (be2aecd)
• add "return 1" to the function "warnmsg" (148c100)
• a little code refactoring (f68df54, 9a93308, 82dc0fd, c2d5d53)

v1.1

(4fb3897) Connect via unix socket (by default)
(ddc5f28) Ignore pg_catalog indexes

v1.0

Initial release to streamline generating downstream releases.