Diego Salvi

Failed to restart cluster with old "untouched" tablespaces

2

**BUG REPORT** - Please describe the issue you observed: I've a long running cluster with many tablespaces. Some of them aren't really used (no transactions, no datachange). They are checkpointed...

Irrecoverable truncate on BRIN

1

**BUG REPORT** 1. Please describe the issue you observed: When applying a truncate on BRIN indexes every data is immediately deleted. If the tablespace is restarted before the next checkpoint...

ReadOnly mode

1

**FEATURE REQUEST** 1. Provide a way to start HerdDB in read-only mode. HerdDB should not modify data pages (and index pages) in any way. No writes, no deletes, no checkpoints,...

Currently object size for page sizing is evaluate statically at compile time. Object size isn't really static and can change between JVMs. Add JOL lib and use it to populate...

Range scans on Indexes doesn't take advantage of index order

1

Queries like `SELECT * FROM mytable WHERE indexedfield >= v ORDER BY indexedfield LIMIT n` ideally should scan just n values. Currently every value after v is retrieved, filtered and...

Activator too "active"

1

When is not possible to start a tablespace activator will keep to try ever and ever again eventually leading to file system full with logs. For instance this error kept...

Actually read or writes takes an exclusive lock on blocks. We could change it to allow multiple readers at once per block.

Checkpoint: Reorder operations

1

Actually checkpoint execute page manipulations as: cleanup dirty page 1. cleanup dirty pages 2. compact small pages 3. flush temporary pages Just step 3 is strictly needed. The process should...

We should add a way to choose page sizing a collaborative way between paged "modules" (blink, brin, tables). Each module have and know his byte overhead per page and it...

BRIN indexes don't take in account fixed block size in overall page size nor map entries. We should add such evaluations like done on BLink