apache / apache/iotdb

CatchupTask may cause a cluster crash?

未关闭
#4,352 3 条评论 0 个 reaction 已指派 0 人 在 GitHub 查看
主要语言
Java
星标
6.4k
派生
1.2k
平均合并
1 天 23 小时
30 天内合并 PR
115

描述

# Description
When my environment(3 nodes,3 replicas) has network fluctuations, or a node is overloaded and responds slowly, after it rejoined the cluster,i find that `CatchupTask` may cause my cluster to be corrupted. So I did some analysis and found the following. Please correct me if there is anything wrong.
# Question
1. `CatchupTask` does not control the size of data on a single slot. In another word,it can result in too many schema or files on a slot _(like slot[981] and slot[911])_,this could be a heavy operation. Besides, since we limit the maximum size of thrift frame to 512 MB,that means the request can not be sent to another node successfully.
```
2021-11-05 17:39:13,148 [DataClientThread-133] INFO o.a.i.c.s.m.DataGroupMember:396 - Data(x.x.x.x:9003, raftId=0): received a snapshot from RaftNode(node:Node(internalIp:x.x.x.x, metaPort:9003, nodeIdentifier:-206505346, dataPort:40010, clientPort:6667, clientIp:x.x.x.x), raftId:0) with size 285120061
2021-11-05 17:39:15,093 [DataClientThread-133] INFO o.a.i.c.l.s.PartitionedSnapshot$Installer:165 - Data(x.x.x.x:9003, raftId=0): start to install a snapshot of 3948175-98
2021-11-05 17:39:15,098 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 9 series, index-term: 0-0} into slot[93]
2021-11-05 17:39:15,098 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,110 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 93 is ready
2021-11-05 17:39:15,110 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[121]
2021-11-05 17:39:15,110 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,116 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 121 is ready
2021-11-05 17:39:15,116 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 16 series, index-term: 0-0} into slot[153]
2021-11-05 17:39:15,117 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,120 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 153 is ready
2021-11-05 17:39:15,121 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[160]
2021-11-05 17:39:15,121 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,124 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 160 is ready
2021-11-05 17:39:15,125 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 660 series, index-term: 0-0} into slot[363]
2021-11-05 17:39:15,139 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,143 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 363 is ready
2021-11-05 17:39:15,143 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 1615 series, index-term: 0-0} into slot[366]
2021-11-05 17:39:15,172 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,176 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 366 is ready
2021-11-05 17:39:15,176 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 2768 series, index-term: 0-0} into slot[574]
2021-11-05 17:39:15,227 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,229 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 574 is ready
2021-11-05 17:39:15,230 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[711]
2021-11-05 17:39:15,230 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,232 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 711 is ready
2021-11-05 17:39:15,232 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 309 series, index-term: 0-0} into slot[843]
2021-11-05 17:39:15,239 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,240 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 843 is ready
2021-11-05 17:39:15,241 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 23 series, index-term: 0-0} into slot[879]
2021-11-05 17:39:15,241 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:15,243 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 879 is ready
2021-11-05 17:39:15,243 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 300511 series, index-term: 0-0} into slot[911]
2021-11-05 17:39:20,456 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:202 - Schemas in snapshot are registered
2021-11-05 17:39:20,458 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:304 - Data(x.x.x.x:9003, raftId=0): slot 911 is ready
2021-11-05 17:39:20,459 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{1 files, 1573281 series, index-term: 0-0} into slot[981]
```
![image](https://user-images.githubusercontent.com/44458757/141046373-b15f220d-3b7b-4097-8eef-04f83da61b6b.png)

When a slot is blocked in a request, the request fails and is retried repeatedly.

At the same time, as the operation increases, the request becomes larger and will never be successfully executed. These threads are taking up resources and the number of threads is increasing.
```
2021-11-05 17:39:15,098 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 9 series, index-term: 0-0} into slot[93]
2021-11-05 17:39:15,110 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[121]
2021-11-05 17:39:15,116 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 16 series, index-term: 0-0} into slot[153]
2021-11-05 17:39:15,121 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[160]
2021-11-05 17:39:15,125 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 660 series, index-term: 0-0} into slot[363]
2021-11-05 17:39:15,143 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 1615 series, index-term: 0-0} into slot[366]
2021-11-05 17:39:15,176 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 2768 series, index-term: 0-0} into slot[574]
2021-11-05 17:39:15,230 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[711]
2021-11-05 17:39:15,232 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 309 series, index-term: 0-0} into slot[843]
2021-11-05 17:39:15,241 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 23 series, index-term: 0-0} into slot[879]
2021-11-05 17:39:15,243 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 300511 series, index-term: 0-0} into slot[911]
2021-11-05 17:39:20,459 [DataClientThread-133] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{1 files, 1573281 series, index-term: 0-0} into slot[981]

2021-11-05 19:38:41,003 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 9 series, index-term: 0-0} into slot[93]
2021-11-05 19:38:41,006 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[121]
2021-11-05 19:38:41,007 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 16 series, index-term: 0-0} into slot[153]
2021-11-05 19:38:41,010 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[160]
2021-11-05 19:38:41,012 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 660 series, index-term: 0-0} into slot[363]
2021-11-05 19:38:41,095 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 1615 series, index-term: 0-0} into slot[366]
2021-11-05 19:38:41,300 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 2768 series, index-term: 0-0} into slot[574]
2021-11-05 19:38:41,684 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[711]
2021-11-05 19:38:41,686 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 309 series, index-term: 0-0} into slot[843]
2021-11-05 19:38:41,726 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 23 series, index-term: 0-0} into slot[879]
2021-11-05 19:38:41,730 [DataClientThread-216] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 300511 series, index-term: 0-0} into slot[911]

2021-11-05 19:39:54,016 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 9 series, index-term: 0-0} into slot[93]
2021-11-05 19:39:54,020 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[121]
2021-11-05 19:39:54,022 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 16 series, index-term: 0-0} into slot[153]
2021-11-05 19:39:54,025 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[160]
2021-11-05 19:39:54,030 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 660 series, index-term: 0-0} into slot[363]
2021-11-05 19:39:54,337 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 1615 series, index-term: 0-0} into slot[366]
2021-11-05 19:39:54,535 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 2768 series, index-term: 0-0} into slot[574]
2021-11-05 19:39:54,876 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 4 series, index-term: 0-0} into slot[711]
2021-11-05 19:39:54,878 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 309 series, index-term: 0-0} into slot[843]
2021-11-05 19:39:54,928 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 23 series, index-term: 0-0} into slot[879]
2021-11-05 19:39:54,933 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 300511 series, index-term: 0-0} into slot[911]
2021-11-05 19:40:43,473 [DataClientThread-218] INFO o.a.i.c.l.s.FileSnapshot$Installer:200 - Starting to install a snapshot FileSnapshot{0 files, 1573281 series, index-term: 0-0} into slot[981]
```

2. When a node is restarted, the sequence of `local recovery` and `peer recovery` is not controlled. `local recovery` could be slow due to a large `mlog.bin`,but `peer recovery` has begun. Although I haven't find what was wrong with it, it was obvious that the CPU load was climbing and the log files were reporting a lot of errors(Because `CatchupTask` has restored the schema, the `local recovery` starts repeating these operations).

![image](https://user-images.githubusercontent.com/44458757/141048390-a51c295b-d147-4dc4-96a1-cc77db8b92e6.png)

# Some thinking
Maybe, In `catchupTask`, we need to control both the size of a request and the amount of data on a single slot.
Assuming that the number of schema on a single slot is one million, we might need to split it into 10 or more operations. Also, we need to control the size of the entire request to ensure that it does not exceed the thrift Frame limit (512MB).

In addition, when a node is restarted, the `local recovery` should be prioritized. The `peer recovery` can start only after the local recovery is complete. And we need to consider whether the conditions for `mtree-snapshot` are too strict, If no snapshot is taken for a long time, the `mlog.bin` file is too large and the recovery speed of nodes in the cluster is inconsistent, which may cause other problems(For example,nodes with slow recovery speed cannot be connected with others, repeated operations during recovery, and so on).

WDYT?

贡献指南

打开贡献指南

调研方向

从 CatchupTask 开始,跟踪经过 DataGroupMember、PartitionedSnapshot.Installer 和 FileSnapshot.Installer 的 snapshot 流程,并以记录的 slot 911 和 slot 981 案例作为切入点。复现一个存在延迟或重新加入的节点,并制造网络波动,然后验证重试不会增加 requests 或 threads 的数量,且 catch-up 能够完成而不会损坏 cluster。

由索引模型根据 Issue 内容生成。

评估

技术栈
java
领域
databases, distributed-systems
Issue 类型
缺陷
难度
4/5
预计耗时
3-5 天
活跃度
停滞
描述清晰度
需要澄清
新手友好度
25/100

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。