Until the spider directives are moved out of the COMMENT, perhaps something like this would be okay?

Create Table: CREATE TABLE `t1` (
`id` bigint(20) unsigned NOT NULL AUTO_INCREMENT,
`val` varchar(255) CHARACTER SET utf8mb4 DEFAULT NULL,
PRIMARY KEY (`id`)
) ENGINE=SPIDER COMMENT='wrapper "mysql", table "t1" CREATE IF NOT EXIST'
PARTITION BY RANGE (`id`)
(PARTITION `p0` VALUES LESS THAN (1000000000) COMMENT = 'srv "spider1node1"' ENGINE = SPIDER,
PARTITION `p1` VALUES LESS THAN (2000000000) COMMENT = 'srv "spider1node2"' ENGINE = SPIDER,
PARTITION `p2` VALUES LESS THAN (3000000000) COMMENT = 'srv "spider1node3"' ENGINE = SPIDER,
PARTITION `pMax` VALUES LESS THAN MAXVALUE COMMENT = 'srv "spider1nodeMax"' ENGINE = SPIDER)

This is a difficult problem to solve because DDLs are non-transactional. If there are any failures on any of the data nodes, then Spider would need to do cleanup to restore consistency because there is no rollback for it.

Yes, until we get transactional DDL, indeed an /excellent/ solution is not likely to exist.

However, what we have today is certainly not a /better/ situation: consider if someone does an alter table, things can just break anyway.

I'd rather have some syntax – even if not the default – which creates the situation where DBAs step-in only when something unexpected goes wrong, rather than for every ADD COLUMN.

It's hard for me to imagine using Spider widely in production without some form of this.