【详述】主键模型实时写入数据，数据量：几千条/s，checkpoint 30s，表数据总量目前6亿，按天分区，Be和Fe都有大量告警，告警信息如下，这个是因为写入压力太大，通过调参解决吗？这个写入速率正常不用调参应该也可以支撑吧
【StarRocks版本】3.0.1
【集群规模】3fe（2 follower）+5be
【机器信息】16C64G
【表结构信息】(大概10个字段)
CREATE TABLE table_name (
id bigint(20) NOT NULL COMMENT “id”,
add_time datetime NOT NULL COMMENT “add_time”
) ENGINE=OLAP
PRIMARY KEY(id, add_time)
PARTITION BY date_trunc(‘day’, add_time)
DISTRIBUTED BY HASH(id)
PROPERTIES (
“replication_num” = “3”,
“enable_persistent_index” = “true”,
“partition_live_number” = “366”,
);
【告警信息】
be.warning
publish_version.cpp:131] publish version threadpool is busy, retry in 50ms. txn_id: 27755595, tablet:2361154
fe.warning
WARN (ReportHandler|87) [TabletInvertedIndex.tabletReport():194] replica 2347164 of tablet 2347162 on backend 11002 need recovery. replica in FE: [replicaId=2347164, BackendId=11002, version=183334, versionHash=0, dataSize=71630423, rowCount=1059947, lastFailedVersion=-1, lastFailedVersionHash=0, lastSuccessVersion=183334, lastSuccessVersionHash=0, lastFailedTimestamp=-1, schemaHash=799523325, minReadableVersion=182628, state=NORMAL], report version 183335, report schema hash: 799523325, is bad: unknown, is version missing: true