Reorganize statement after cluster update seems to hang
Details
| Detail name | Value |
|---|---|
| Changelog Number | 7494 |
| Type | Bug |
| Status | Resolved |
| Affected Versions | Exasol 6.1.1 |
| Fix Versions | Exasol 6.1.2 |
| Resolution Date | 2019-02-15 |
Synopsis
If some specific circumstances are met (see below), for some tables the first reorganize statement after a cluster reorganize does not finish in a reasonable time leading to the perception of it being deadlocked.
We stress that this bug occurs only for the first reorganize statement after cluster enlargement, further reorganize statements run normally.
Details
For this bug to occur, the following conditions must be met:
- table is not too small
- no distribution keys are set on the table
- many deleted rows (internally they need to be spread out across the table, i.e. not a few chunks of consecutive deleted rows)
If these requirements are met, the first reorganize statement after a cluster enlargement may run in a performance bug which make it run for a very long time.
Workaround
As a workaround, we recommend to set a temporary distribution key:
alter table <table_name> distribute by <some column>; alter table <table_name> drop distribution keys;
