Spider has so many plugin variables and some (most) of them result in excessive tweak. Deprecate such variables. We also need to deprecate the corresponding table parameters.
TODO:
List all variables to be deprecated
Check the list again (bolded are checked)- Deprecate them one by one
Docs:
- https://mariadb.com/kb/en/spider-server-system-variables/
- https://tendbcluster.com/book-en/Documentation/re-book/tspider-parameter-en.html
x: deprecate, o: do not deprecate (at least now), empty: postpone judgment
| Deprecate | Variable Name | Comment |
|---|---|---|
| o | spider_auto_increment_mode | |
| x | spider_bgs_first_read | The best value should be specified by the developers. |
| spider_bgs_mode | ||
| x | spider_bgs_second_read | The best value should be specified by the developers. |
| x | spider_bka_engine | Always use MEMORY. |
| x | spider_bka_mode | |
| spider_bka_table_name_type | Flexibility on the table name is not necessary. | |
| MDEV-28877 | spider_block_size | There is no need to change the block size. |
| spider_buffer_size | This is used in nowhere in the codebase. | |
| o | spider_bulk_size | Merge to spider_bulk_write_size |
| x | spider_bulk_update_mode | Always use bulk update/delete. |
| o | spider_bulk_update_size | Merge to spider_bulk_size |
| x | spider_casual_read | Insignificant but lot of complication but not fully sure. The value should be 1. |
| o | spider_conn_recycle_mode | One might want to disconnect every time. For example, for DNS based HA solutions. |
| x | spider_conn_recycle_strict | Looks not so useful. |
| o | spider_conn_wait_timeout | |
| x | spider_connect_error_interval | Do not do this kind of back off at Spider. |
| x | spider_connect_mutex | No one use this. |
| MDEV-27925 | spider_connect_retry_count | Do not retry if it got too many connections. |
| MDEV-27925 | spider_connect_retry_interval | Do not retry if it got too many connections. |
| o | spider_connect_timeout | |
| MDEV-27253 | spider_crd_bg_mode | |
| MDEV-27253 | spider_crd_interval | |
| spider_crd_mode | ||
| MDEV-27253 | spider_crd_sync | |
| spider_crd_type | The best value should be choosen by the engine developer. Or, adjusted automatically. | |
| spider_crd_weight | The best value should be choosen by the engine developer. Or, adjusted automatically. | |
| spider_delete_all_rows_type | Specify whether to use ha_spider::delete_all_rows() or not. Always use delte_all_rows(). | |
| x | spider_direct_dup_insert | Always do check on remote nodes. |
| x | spider_direct_order_limit | Always pushdown ORDER BY and LIMIT. The documentation is wrong. |
| o | spider_dry_access | https://github.com/Tencent/TenDBCluster-TSpider/commit/f0860e2b7d04582a5d00130b2963728ff4d3f9f9 |
| x | spider_error_read_mode | How does this help users? > "Useful with applications that don't implement transaction replays." |
| x | spider_error_write_mode | The same as the above. |
| x | spider_first_read | The best value should be specified by the engine developers. |
| o | spider_force_commit | Not a good name. |
| o | spider_general_log | |
| o | spider_index_hint_pushdown | |
| spider_init_sql_alloc_size | This shouldn't be exposed to users. | |
| spider_internal_limit | This doesn't look useful rather dangerous. | |
| spider_internal_offset | This doesn't look useful rather dangerous. | |
| MDEV-28630 | spider_internal_optimize | Spider doesn't pushes down DDLs, so it seems to be natural not to pushdown table maintenance statements too. |
| MDEV-28630 | spider_internal_optimize_local | This should be merged to spider_internal_optimize at least or just deprecate. |
| x | spider_internal_sql_log_off | This seems to be for testing. Removing the vraible may affects the tests so much. In such a case, create a separated JIRA issue. |
| x | spider_internal_unlock | This specify whether Spider pass through UNLOCK TABLE to data nodes or not. I assume this is for internal use. |
| o | spider_internal_xa | |
| x | spider_internal_xa_id_type | This should be specified by the engine developer. |
| spider_internal_xa_snapshot | ||
| spider_load_crd_at_startup | ||
| spider_load_sts_at_startup | ||
| MDEV-28564 | spider_local_lock_table | Do not use local locks. |
| x | spider_lock_exchange | How does this help users? |
| o | spider_log_result_error_with_sql | Merge to spider_log_level |
| o | spider_log_result_errors | Merge to spider_log_level |
| x | spider_low_mem_read | Avoid branching. |
| o | spider_max_connections | |
| x | spider_max_order | When to use this? |
| o | spider_multi_split_read | |
| o | spider_net_read_timeout | The default (table) value 600 seconds is too long and should be changed. |
| o | spider_net_write_timeout | The default (table) value 600 seconds is too long and should be changed. |
| o | spider_ping_interval_at_trx_start | |
| x | spider_quick_mode | This variable might be deprecated too. |
| x | spider_quick_page_byte | |
| x | spider_quick_page_size | |
| o | spider_read_only_mode | |
| o | spider_remote_access_charset | |
| o | spider_remote_autocommit | |
| o | spider_remote_default_database | |
| o | spider_remote_sql_log_off | |
| o | spider_remote_time_zone | |
| o | spider_remote_trx_isolation | This might be able to be merged with spider_sync_trx_isolation |
| o | spider_remote_wait_timeout | |
| MDEV-27228 | spider_reset_sql_alloc | I don't know why this is exposed to users. |
| o | spider_same_server_link | |
| x | spider_second_read | The best value should be specified by the engine developers. |
| x | spider_select_column_mode | The best mode should be chosen by the engine developer. Here, the value should be 1. |
| x | spider_selupd_lock_mode | Setting local locks makes sense only when only a single Spider node is active. |
| x | spider_semi_split_read | https://tendbcluster.com/book-en/Documentation/re-book/tspider-parameter-en.html |
| x | spider_semi_split_read_limit | https://tendbcluster.com/book-en/Documentation/re-book/tspider-parameter-en.html |
| spider_semi_table_lock | This result in poor performance. Almost everyone use transactional storage engines. | |
| spider_semi_table_lock_connection | ||
| x | spider_semi_trx | We deprecate this because semi-split read is depreacted too. |
| x | spider_semi_trx_isolation | Very bad name. |
| x | spider_skip_default_condition | Specify whether to skip the default case of the switch-case statement in open_item_func(). This looks a kind of a internal variable. Further, the default case should be always skiped for the safety. The variable is also used in the group by handler but it is not obvious how it affects the storage engine behavior. Thus, we do not delete this at this time. |
| o | spider_skip_parallel_search | Introduced by MariaDB 10.3.3. TSpider's default value is 3. |
| o | spider_slave_trx_isolation | Recently introduced by |
| x | spider_split_read | Not well-documented. |
| spider_store_last_crd | ||
| spider_store_last_sts | ||
| x | spider_strict_group_by | Always use strictly. |
| MDEV-27253 | spider_sts_bg_mode | |
| MDEV-27253 | spider_sts_interval | |
| spider_sts_mode | ||
| MDEV-27253 | spider_sts_sync | |
| spider_support_xa | Always support XA. | |
| MDEV-28847 | spider_sync_autocommit | |
| x | spider_sync_sql_mode | Sync always. |
| o | spider_sync_trx_isolation | Not sure but might be useful when working with RDBMSs other than MariaDB. |
| spider_table_crd_thread_count | ||
| MDEV-28563 | spider_table_init_error_interval | Spider should behave in the way of the variable is set to 0? |
| spider_table_sts_thread_count | ||
| spider_udf_ct_bulk_insert_interval | Use optional argument of the UDF. | |
| spider_udf_ct_bulk_insert_rows | Use optional argument of the UDF. | |
| spider_udf_ds_bulk_insert_rows | Use optional argument of the UDF. | |
| spider_udf_ds_table_loop_mode | ||
| spider_udf_ds_use_real_table | ||
| spider_udf_table_lock_mutex_count | Looks rather an internal variable. | |
| spider_udf_table_mon_mutex_count | Looks rather an internal variable. | |
| MDEV-27228 | spider_use_all_conns_snapshot | This complicates the connection management. |
| x | spider_use_cond_other_than_pk_for_update | Specify whether to append non-primary key search conditions to a generated query. The variable seems to have been introduced for a temporal workaround for |
| MDEV-28794 | spider_use_consistent_snapshot | |
| o | spider_use_default_database | |
| x | spider_use_flash_logs | Spider doesn't pushes down DDLs, so it seems to be natural not to pushdown table maintenance statements too. |
| spider_use_handler | Converting SQL statements to HANDLER statements may improve performance in some case. However, the performance gain seems not to be very significant and this complicates the implementation. Setting spider_use_handler > 0 disables the GROUP BY handler. | |
| spider_use_pushdown_udf | ||
| MDEV-28792 | spider_use_snapshot_with_flush_tables | This complicates the connection management. |
| x | spider_use_table_charset | Always use table charset. |
| o | spider_version | |
| o | spider_wait_timeout | |
| spider_xa_register_mode | The variable only affects the internal XA transactions. |