etcd: Inconsistent Revisions Across Members ( v3.3.3 )
Etcd Version: 3.3.3 OS: Linux 14.04.1 Ubuntu
Health
http://10.213.214.4:2379 is healthy: successfully committed proposal: took = 743.509µs
http://10.213.214.2:2379 is healthy: successfully committed proposal: took = 1.00709ms
http://10.213.214.3:2379 is healthy: successfully committed proposal: took = 1.011886ms
Table output
+--------------------------+------------------+---------+---------+-----------+-----------+------------+
| ENDPOINT | ID | VERSION | DB SIZE | IS LEADER | RAFT TERM | RAFT INDEX |
+--------------------------+------------------+---------+---------+-----------+-----------+------------+
| http://10.213.214.2:2379 | 372c0ce563084c78 | 3.3.3 | 95 MB | false | 18 | 12165951 |
| http://10.213.214.3:2379 | 8546c864475e6358 | 3.3.3 | 136 MB | false | 18 | 12165952 |
| http://10.213.214.4:2379 | e030b17253d751e9 | 3.3.3 | 175 MB | true | 18 | 12165953 |
+--------------------------+------------------+---------+---------+-----------+-----------+------------+
JSON Ouput
[
{
"Endpoint":"http://10.213.214.2:2379",
"Status": {
"header {
"cluster_id":16002117407141323128,"
"member_id":3975566750409837688,
"revision":1377021,
"raft_term":18
},
"version":"3.3.3",
"dbSize":95182848,
"leader":16154606967968977385,
"raftIndex":12165645,
"raftTerm":18
}
},
{
"Endpoint":"http://10.213.214.3:2379",
"Status":{
"header":{
"cluster_id":16002117407141323128,
"member_id":9603583588433814360,
"revision":1801804,
"raft_term":18
},
"version":"3.3.3",
"dbSize":136417280,
"leader":16154606967968977385,
"raftIndex":12165646,
"raftTerm":18
}
},
{
"Endpoint":"http://10.213.214.4:2379",
"Status":{
"header"{
"cluster_id":16002117407141323128,
"member_id":16154606967968977385,
"revision":891479,
"raft_term":18
},
"version":"3.3.3",
"dbSize":174526464,
"leader":16154606967968977385,
"raftIndex":12165647,
"raftTerm":18
}
}
]
As you can see the revisions across members are vastly different.
Each of these nodes are configured with --auto-compaction-retention 1
.
Any thoughts on what’s going on here?
About this issue
- Original URL
- State: closed
- Created 5 years ago
- Comments: 17 (14 by maintainers)
It is concerning that there is such difference between
revision
on different members. Can you check metricsbackend_commit_duration_seconds
on each node?Can you provide more details on this?
In etcd, difference in
revision
does not mean inconsistency in key value result. Member with lowerrevision
cannot serve latest read request (assuming you are using linearizable read, which is the default, see [1]) until its backend catches up.[1] https://github.com/etcd-io/etcd/blob/952b9e75c65f7c1f5490511610293a127131608e/etcdserver/etcdserverpb/rpc.proto#L414-L420