storaged不断升高后宕机

  • nebula 版本:3.3.0
  • 部署方式:分布式
  • 安装方式:RPM
  • 是否为线上版本:N
  • 硬件信息
    • 磁盘 (SSD两台,机械三台)
    • CPU、内存信息(两台64核128G内存 三台32核 48G)
  • 问题的具体描述
    当前数据量点约为4亿 边约为8亿 使用flink导入边时报错,发现storage一台节点宕机,查看dashboard发现该节点内存从部署开始不断升高 (该节点为48G内存),于今早6点到达顶峰46G 后宕机

日志如下

I20221208 06:13:45.560936 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 21 files into 0, base level is 0, output level is 1
I20221208 06:13:45.560997 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:13:57.699807 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 21 files into 5, base level is 0, output level is 1
I20221208 06:13:57.736078 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelMaxLevelSize, status: OK, compacted 11 files into 0, base level is 1, output level is 2
I20221208 06:13:57.736140 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:14:23.799724 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelMaxLevelSize, status: OK, compacted 11 files into 11, base level is 1, output level is 2
I20221208 06:14:58.078831 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 0, base level is 0, output level is 1
I20221208 06:14:58.078939 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:15:09.243021 52669 EventListener.h:158] Stall conditions changed column family: default, current condition: Delayed, previous condition: Normal
I20221208 06:15:11.118047 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 4, base level is 0, output level is 1
I20221208 06:15:11.146811 52668 EventListener.h:158] Stall conditions changed column family: default, current condition: Normal, previous condition: Delayed
I20221208 06:15:11.146991 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 20 files into 0, base level is 0, output level is 1
I20221208 06:15:11.147042 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:15:23.681516 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 20 files into 5, base level is 0, output level is 1
I20221208 06:15:23.867177 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelMaxLevelSize, status: OK, compacted 10 files into 0, base level is 1, output level is 2
I20221208 06:15:23.867264 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:15:50.560184 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelMaxLevelSize, status: OK, compacted 10 files into 10, base level is 1, output level is 2
I20221208 06:15:50.651342 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelMaxLevelSize, status: OK, compacted 7 files into 0, base level is 2, output level is 3
I20221208 06:15:50.652829 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:16:11.169066 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelMaxLevelSize, status: OK, compacted 7 files into 6, base level is 2, output level is 3
I20221208 06:16:14.218916 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 0, base level is 0, output level is 1
I20221208 06:16:14.219014 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:16:45.254184 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 4, base level is 0, output level is 1
I20221208 06:16:45.288691 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 19 files into 0, base level is 0, output level is 1
I20221208 06:16:45.291432 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:16:56.227257 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 19 files into 4, base level is 0, output level is 1
I20221208 06:17:29.164124 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 0, base level is 0, output level is 1
I20221208 06:17:29.164626 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:17:40.122253 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 4, base level is 0, output level is 1
I20221208 06:17:50.487905 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 0, base level is 0, output level is 1
I20221208 06:17:50.489799 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:18:17.771726 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 8 files into 5, base level is 0, output level is 1
I20221208 06:18:17.811136 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 18 files into 0, base level is 0, output level is 1
I20221208 06:18:17.811203 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:18:32.787885 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 18 files into 6, base level is 0, output level is 1
I20221208 06:18:32.831928 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelMaxLevelSize, status: OK, compacted 9 files into 0, base level is 1, output level is 2
I20221208 06:18:32.831980 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:18:55.472051 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelMaxLevelSize, status: OK, compacted 9 files into 9, base level is 1, output level is 2
I20221208 06:18:55.528242 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelMaxLevelSize, status: OK, compacted 5 files into 0, base level is 2, output level is 3
I20221208 06:18:55.528296 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:19:07.416523 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelMaxLevelSize, status: OK, compacted 5 files into 4, base level is 2, output level is 3
I20221208 06:19:07.466044 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelMaxLevelSize, status: OK, compacted 7 files into 0, base level is 2, output level is 3
I20221208 06:19:07.466107 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:19:33.903349 52669 EventListener.h:158] Stall conditions changed column family: default, current condition: Delayed, previous condition: Normal
I20221208 06:19:40.647406 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelMaxLevelSize, status: OK, compacted 7 files into 7, base level is 2, output level is 3
I20221208 06:19:40.705735 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelL0FilesNum, status: OK, compacted 25 files into 0, base level is 0, output level is 1
I20221208 06:19:40.705837 52668 CompactionFilter.h:92] Do default minor compaction!
I20221208 06:19:54.354334 52668 EventListener.h:35] Rocksdb compaction completed column family: default because of LevelL0FilesNum, status: OK, compacted 25 files into 5, base level is 0, output level is 1
I20221208 06:19:54.392439 52668 EventListener.h:158] Stall conditions changed column family: default, current condition: Normal, previous condition: Delayed
I20221208 06:19:54.392637 52668 EventListener.h:21] Rocksdb start compaction column family: default because of LevelMaxLevelSize, status: OK, compacted 10 files into 0, base level is 1, output level is 2
I20221208 06:19:54.392685 52668 CompactionFilter.h:92] Do default minor compaction!

应该是因为OOM被系统kill了,可以通过dmesg确认一下。至于解决方法,能否先贴一下storaged.conf看一下配置?

论坛类似建议很多,大体就是把rocksdb的各种占用减少

此话题已在最后回复的 30 天后被自动关闭。不再允许新回复。