存算一体 add column 触发 be节点内存瞬间暴增直到挂掉

💬 StarRocks 用户问答 功能使用相关
#1

【详述】存算一体 在给表增加字段时,触发 be节点内存瞬间暴增直到挂掉

  1. 动态分区聚合表
  2. 存储一个月的数据, 天级分区
  3. 总大小 700多 G
  4. 对表执行 增加列操作
ALTER TABLE rtb.rtb_oks_technology ADD COLUMN material_id int KEY AFTER creative_id;
  1. 增加失败, 通过
SHOW ALTER TABLE COLUMN FROM rtb ORDER BY createtime DESC LIMIT 20;

查看结果提示异常

schema change task failed: 
task type: ALTER, 
status_code: RUNTIME_ERROR, 
alter failed,
status: STARROCKS_ERROR,Memory of process exceed limit. read and decompress page Backend: rtb-sr-be05, Used: 28057267632, Limit: 25510835957. Mem usage has exceed the limit of BE, backendId: 10006, signature: 911367
  1. 查看be日志, 所有的 be 节点全部挂掉, 我们是 systemctl 配置了自启动, be 节点在挂掉后重启了

【业务影响】
【是否存算分离】 否
【StarRocks版本】 3.3.5/3.3.6 都存在此问题
【集群规模】3fe + 6be
【机器信息】16核 32G 内存
【联系方式】
【附件】
表结构

CREATE TABLE `rtb_oks_technology` (
  `__d` date NOT NULL COMMENT "",
  `__time` int(11) NOT NULL COMMENT "",
  `tid` int(11) NULL COMMENT "",
  `media_bundle` varchar(256) NULL COMMENT "",
  `position_id` varchar(256) NULL COMMENT "",
  `country` varchar(3) NULL COMMENT "",
  `tml_play_order` tinyint(4) NULL COMMENT "",
  `h` varchar(10) NULL COMMENT "",
  `ad_size` varchar(32) NULL COMMENT "",
  `adx_id` int(11) NULL COMMENT "",
  `adm_tech` int(11) NULL COMMENT "",
  `creative_id` int(11) NULL COMMENT "",
  `tml_play` bigint(20) SUM NULL COMMENT "",
  `impression` bigint(20) SUM NULL COMMENT "adt impression",
  `adt_impression` bigint(20) SUM NULL COMMENT "",
  `burl` bigint(20) SUM NULL COMMENT ""
) ENGINE=OLAP 
AGGREGATE KEY(`__d`, `__time`, `tid`, `media_bundle`, `position_id`, `country`, `tml_play_order`, `h`, `ad_size`, `adx_id`, `adm_tech`, `creative_id`)
COMMENT "OLAP"
PARTITION BY RANGE(`__d`)
(PARTITION p20241001 VALUES [("2024-10-01"), ("2024-10-02")),
PARTITION p20241002 VALUES [("2024-10-02"), ("2024-10-03")),
PARTITION p20241003 VALUES [("2024-10-03"), ("2024-10-04")),
PARTITION p20241004 VALUES [("2024-10-04"), ("2024-10-05")),
PARTITION p20241005 VALUES [("2024-10-05"), ("2024-10-06")),
PARTITION p20241006 VALUES [("2024-10-06"), ("2024-10-07")),
PARTITION p20241007 VALUES [("2024-10-07"), ("2024-10-08")),
PARTITION p20241008 VALUES [("2024-10-08"), ("2024-10-09")),
PARTITION p20241009 VALUES [("2024-10-09"), ("2024-10-10")),
PARTITION p20241010 VALUES [("2024-10-10"), ("2024-10-11")),
PARTITION p20241011 VALUES [("2024-10-11"), ("2024-10-12")),
PARTITION p20241012 VALUES [("2024-10-12"), ("2024-10-13")),
PARTITION p20241013 VALUES [("2024-10-13"), ("2024-10-14")),
PARTITION p20241014 VALUES [("2024-10-14"), ("2024-10-15")),
PARTITION p20241015 VALUES [("2024-10-15"), ("2024-10-16")),
PARTITION p20241016 VALUES [("2024-10-16"), ("2024-10-17")),
PARTITION p20241017 VALUES [("2024-10-17"), ("2024-10-18")),
PARTITION p20241018 VALUES [("2024-10-18"), ("2024-10-19")),
PARTITION p20241019 VALUES [("2024-10-19"), ("2024-10-20")),
PARTITION p20241020 VALUES [("2024-10-20"), ("2024-10-21")),
PARTITION p20241021 VALUES [("2024-10-21"), ("2024-10-22")),
PARTITION p20241022 VALUES [("2024-10-22"), ("2024-10-23")),
PARTITION p20241023 VALUES [("2024-10-23"), ("2024-10-24")),
PARTITION p20241024 VALUES [("2024-10-24"), ("2024-10-25")),
PARTITION p20241025 VALUES [("2024-10-25"), ("2024-10-26")),
PARTITION p20241026 VALUES [("2024-10-26"), ("2024-10-27")),
PARTITION p20241027 VALUES [("2024-10-27"), ("2024-10-28")),
PARTITION p20241028 VALUES [("2024-10-28"), ("2024-10-29")),
PARTITION p20241029 VALUES [("2024-10-29"), ("2024-10-30")),
PARTITION p20241030 VALUES [("2024-10-30"), ("2024-10-31")),
PARTITION p20241031 VALUES [("2024-10-31"), ("2024-11-01")),
PARTITION p20241101 VALUES [("2024-11-01"), ("2024-11-02")),
PARTITION p20241102 VALUES [("2024-11-02"), ("2024-11-03")),
PARTITION p20241103 VALUES [("2024-11-03"), ("2024-11-04")),
PARTITION p20241104 VALUES [("2024-11-04"), ("2024-11-05")),
PARTITION p20241105 VALUES [("2024-11-05"), ("2024-11-06")),
PARTITION p20241106 VALUES [("2024-11-06"), ("2024-11-07")),
PARTITION p20241107 VALUES [("2024-11-07"), ("2024-11-08")),
PARTITION p20241108 VALUES [("2024-11-08"), ("2024-11-09")),
PARTITION p20241109 VALUES [("2024-11-09"), ("2024-11-10")),
PARTITION p20241110 VALUES [("2024-11-10"), ("2024-11-11")),
PARTITION p20241111 VALUES [("2024-11-11"), ("2024-11-12")),
PARTITION p20241112 VALUES [("2024-11-12"), ("2024-11-13")),
PARTITION p20241113 VALUES [("2024-11-13"), ("2024-11-14")),
PARTITION p20241114 VALUES [("2024-11-14"), ("2024-11-15")),
PARTITION p20241115 VALUES [("2024-11-15"), ("2024-11-16")),
PARTITION p20241116 VALUES [("2024-11-16"), ("2024-11-17")),
PARTITION p20241117 VALUES [("2024-11-17"), ("2024-11-18")),
PARTITION p20241118 VALUES [("2024-11-18"), ("2024-11-19")),
PARTITION p20241119 VALUES [("2024-11-19"), ("2024-11-20")),
PARTITION p20241120 VALUES [("2024-11-20"), ("2024-11-21")),
PARTITION p20241121 VALUES [("2024-11-21"), ("2024-11-22")),
PARTITION p20241122 VALUES [("2024-11-22"), ("2024-11-23")),
PARTITION p20241123 VALUES [("2024-11-23"), ("2024-11-24")),
PARTITION p20241124 VALUES [("2024-11-24"), ("2024-11-25")),
PARTITION p20241125 VALUES [("2024-11-25"), ("2024-11-26")),
PARTITION p20241126 VALUES [("2024-11-26"), ("2024-11-27")),
PARTITION p20241127 VALUES [("2024-11-27"), ("2024-11-28")),
PARTITION p20241128 VALUES [("2024-11-28"), ("2024-11-29")),
PARTITION p20241129 VALUES [("2024-11-29"), ("2024-11-30")),
PARTITION p20241130 VALUES [("2024-11-30"), ("2024-12-01")),
PARTITION p20241201 VALUES [("2024-12-01"), ("2024-12-02")),
PARTITION p20241202 VALUES [("2024-12-02"), ("2024-12-03")))
DISTRIBUTED BY HASH(`__time`) BUCKETS 16 
PROPERTIES (
"compression" = "LZ4",
"dynamic_partition.buckets" = "16",
"dynamic_partition.enable" = "true",
"dynamic_partition.end" = "2",
"dynamic_partition.history_partition_num" = "8",
"dynamic_partition.prefix" = "p",
"dynamic_partition.start" = "-60",
"dynamic_partition.time_unit" = "DAY",
"dynamic_partition.time_zone" = "UTC",
"fast_schema_evolution" = "true",
"replicated_storage" = "true",
"replication_num" = "3"
);

京ICP备2022026421号-1