[IOTDB-3016] Run flush when graceful stop IoTDB

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Fix Version/s: master branch
• Component/s: Core/Others
• Labels: pull-request-available
• Assignee: Bo Zhang
• Reporter: Bo Zhang
• Created: 26/Apr/22 12:42
• Updated: 09/May/22 04:16
• Resolved: 09/May/22 04:16

Description

When we test  1 IoTDB server who has many storagegroups and high TPS, we found below case:

==========================================================

When we closed this IoTDB instance and then restarted it, the IoTDB spent more than 5 minutes before begining to provide service for clients.

After analyzing,  we found that the reason is :

1) IoTDB server's shutdown cause much WAL logs were kept in Disk.  

2) When IoTDB restart, it will spend much time to recover from much WAL logs.

==============================================

So we plan to add flush action before graceful closing IoTDB process.

 

 

 

Issue Links

• duplicates: IOTDB-2929 [stop-server.sh] Flush is necessary when the database is stopped normally
• links to: GitHub Pull Request #5800