SHOW BACKENDS与show proc '/backends'结果输出不一致

【StarRocks版本】2.3.14
【集群规模】4fe(3 follower+1observer)+4be

SHOW BACKENDS与show proc '/backends’结果输出不一致, SHOW BACKENDS显示TabletNum: 352,但是show proc '/backends’显示TabletNum: 0

SHOW BACKENDS\G结果如下图:


show proc ‘/backends’\G结果如下图:
image

fe master节点的日志有关于be节点1795069的信息
2024-02-22 16:38:43,841 INFO (thrift-server-pool-101|1983) [ReportHandler.handleReport():226] receive report from be 1795069. type: WORKGROUP_REPORT, current queue size: 2
2024-02-22 16:38:44,258 INFO (cluster|17) [SystemHandler.runAlterJobV2():119] backend 1795069 lefts 352 replicas to decommission(show up to 20): [2949574, 2949578, 2949582, 2949586, 2949590, 2949594, 2949598, 2949602, 2949937, 2949941, 2949945, 2949949, 2949953, 2949957, 2949961, 2949965, 2974387, 2974391, 2974395, 2974399]
be id: 1795069, is available: false, mediums: [{medium: HDD, replica: 352, used: 0, total: 6046783791104, score: 9.560904591950371E-4},{medium: SSD, replica: 0, used: 0, total: 0, score: NaN},{medium: S3, replica: 0, used: 0, total: 0, score: NaN},], paths: [{path: /work/starrocks33660/be/storage, path hash: 3025504982510111454, be: 1795069, medium: HDD, used: 0, total: 6046783791104},]
2024-02-22 16:38:48,841 INFO (thrift-server-pool-75|1957) [ReportHandler.handleReport():226] receive report from be 1795069. type: WORKGROUP_REPORT, current queue size: 2
2024-02-22 16:38:52,904 INFO (Thread-43|89) [ReportHandler.taskReport():392] begin to handle task report from backend 1795069
2024-02-22 16:38:52,904 INFO (thrift-server-pool-1|132) [ReportHandler.handleReport():226] receive report from be 1795069. type: TASK_REPORT, current queue size: 1
2024-02-22 16:38:52,905 INFO (Thread-43|89) [ReportHandler.taskReport():432] finished to handle task report from backend 1795069, diff task num: 0. cost: 0 ms
2024-02-22 16:38:53,842 INFO (thrift-server-pool-86|1968) [ReportHandler.handleReport():226] receive report from be 1795069. type: WORKGROUP_REPORT, current queue size: 1
2024-02-22 16:38:54,259 INFO (cluster|17) [SystemHandler.runAlterJobV2():119] backend 1795069 lefts 352 replicas to decommission(show up to 20): [2949574, 2949578, 2949582, 2949586, 2949590, 2949594, 2949598, 2949602, 2949937, 2949941, 2949945, 2949949, 2949953, 2949957, 2949961, 2949965, 2974387, 2974391, 2974395, 2974399]
2024-02-22 16:38:58,842 INFO (thrift-server-pool-101|1983) [ReportHandler.handleReport():226] receive report from be 1795069. type: WORKGROUP_REPORT, current queue size: 2

对be节点1795069进行下线,be节点始终有352个tablet,请问这又是为何?

  • 当天是否做过 删表或者 truncate table 的操作
  • 把fe元数据过期时间调小一点,过段时间再看看 admin set frontend config(“catalog_trash_expire_second”=“300”); (默认1天)

be节点下线后有TRUNCATE TABLE xxxxxx PARTITION (p20240220)的操作,这个会影响吗?

可以设置下上面的参数,对卡住的 be 先取消下线 再重新下线 观察一下