ksarch-saas / cfgserver Goto Github PK

我配置了主主同步，但是从master1修改了key，并没有同步到master2。但是我们用原版的SSDB是可以的，请帮忙分析一下问题，谢谢！

`int Slave::proc_z_clear(const LogEvent &event) {
assert (event.cmd() == BinlogCommand::Z_CLEAR);

Bytes key = event.key();

uint64_t offset = 0;
uint64_t limit = 1000;
int ret = 0;

uint64_t version;
char t = DataType::ZSET;
SLAVE_PROC_CHECK_VERSION(key);
if (t != DataType::ZSET) {
	log_error("unexpected data type: %" PRIu8 " expected:%" PRIu8, uint8_t(t), uint8_t(DataType::ZSET));
	return -1;
}

Transaction trans(serv->ssdb, key);
while (1) {
	ZIterator *it = serv->ssdb->zrange(key, offset, limit, version);
	int num = 0;
	while (it->next()) {
		ret = serv->ssdb->zdel(key, it->key, trans, version);
		if (ret < 0) {
			log_error("zclear zdel failed, type=%" PRId8 " seq=%" PRIu64, event.type(), event.seq());
			num = 0;
			break;
		}
		num++;
	}
	delete it;

	if (num == 0) {
		break;
	}
}

if (ret >= 0 && serv->binlog) {
	serv->binlog->write(BinlogType::SYNC, BinlogCommand::Z_CLEAR, key);
}

return ret;

}`

删除时,zdel应该如下
ret = serv->ssdb->zdel(key, it->field, trans, version);
传it->filed,否则会造成死循环，请更正一下

启动两个ssdb a和b
执行命令
set name a
slot_premigrating 5798
migrate_slot 5798 127.0.0.1 28888 60 1
slot_postmigrating 5798
执行完之后在B节点没有找到该slot的记录，麻烦帮忙看看哪儿出问题了

附上日志
273717 2017-06-13 15:45:43.711 [DEBUG] worker.cpp(35): w:0.048,p:6.159, req: slot_premigrating 5798, resp: ok
273718 2017-06-13 15:46:17.696 [INFO ] range_migrate.cpp(107): range migrating start
273719 2017-06-13 15:46:17.696 [INFO ] range_migrate.cpp(144): connect to 127.0.0.1:28888
273720 2017-06-13 15:46:17.696 [DEBUG] range_migrate.cpp(274): client init
273721 2017-06-13 15:46:17.696 [DEBUG] range_migrate.cpp(172): key migrate init, start: \xff|VERSION|\x00\xa6\x16 end: \xff|VERSION|\x00\xff\xa6\x16
273722 2017-06-13 15:46:17.696 [INFO ] range_migrate.cpp(280): there is no key to sync
273723 2017-06-13 15:46:17.696 [INFO ] range_migrate.cpp(109): range migrating quit, 0 keys sent
273724 2017-06-13 15:46:17.696 [DEBUG] serv.cpp(715): migrate quite
273725 2017-06-13 15:46:17.696 [DEBUG] worker.cpp(35): w:0.020,p:0.437, req: migrate_slot 5798 127.0.0.1 28888 60 1, resp: done
273726 2017-06-13 15:46:30.970 [INFO ] ttl.cpp(46): expiration thread start tid: 139936336439040.
273727 2017-06-13 15:46:30.970 [DEBUG] worker.cpp(35): w:0.040,p:0.178, req: slot_postmigrating 5798, resp: ok

在做大量slot（500个slot）迁移的时候，发现被迁入的节点cpu占用率很高
看代码发现一个slot对应一个线程，
这个能做下优化吗，或者给一个优化的思路。