[jira] [Commented] (IOTDB-1118) [Distributed] Limit query log lag in mid consistency level

["Chao Wang (Jira)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/IOTDB-1118?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17269952#comment-17269952 ] 

Chao Wang commented on IOTDB-1118:
----------------------------------

Thanks [~jt2594838],  your idea is good.

I think,
1. If the number of follower synchronization logs exceeds a specified threshold, the query delivered to followers can be routed to the leader or other normal followers for query. query should be not failed.
2. The leader should be able to detect which nodes exceed the threshold and take corresponding measures to make the follower catch up as soon as possible.

> [Distributed] Limit query log lag in mid consistency level
> ----------------------------------------------------------
>
>                 Key: IOTDB-1118
>                 URL: https://issues.apache.org/jira/browse/IOTDB-1118
>             Project: Apache IoTDB
>          Issue Type: Improvement
>          Components: Cluster
>            Reporter: Tian Jiang
>            Priority: Major
>
> We provide a middle level of read consistency, which is a best-effort strategy, and will try to synchronize with the leader and uses the local data when the synchronization times out. However, the maximum log lag is undefined, which means the node can be arbitrarily far from the leader (missing arbitrarily length of logs) and incurs inconvenience when the user wants to control the data freshness.
> To resolve this, a configuration should be provided like `max_read_log_lag`, which only takes effect when the consistency level is set to mid, and queries will fail if the log lag exceeds this configuration after a time out.



--
This message was sent by Atlassian Jira
(v8.3.4#803005)