The Presto coordinator cannot be started properly.

Issue

The coordinator process of Presto is killed due to an unknown reason, or the coordinator process of Presto cannot be started.

Symptom

The Presto coordinator process cannot be started properly. On the Manager page, it is shown that the presto coordinator process is started properly and its status is normal. However, the background log shows that the coordinator process is not started. Only the following log is displayed:

The Presto coordinator is killed before being started, and no other logs are printed. Other Presto logs do not indicate the reason why the presto coordinator is killed.

Cause Analysis

The port check logic of the presto health check script does not distinguish ports.

Procedure

1. Use a tool to log in to the master nodes of the cluster and perform the following operations:
2. Run the following command to edit the file:

   vim /opt/Bigdata/MRS_xxx/install/FusionInsight-Presto-*/ha/module/harm/plugin/script/pcd.sh

   Change line 31 in the file to http_port_exists=$(netstat -apn | awk '{print $4, $6}' | grep :${HTTP_PORT} | grep LISTEN | wc -l).

   

3. Save the modification. On FusionInsight Manager, choose Services > Presto > Instances to restart the Coordinator process.