Article directory
1. One sentence description
The deployment of starrocks fails on a machine without the AVX2 instruction set. The solution is to replace the machine with the AVX2 instruction set.
Official description:
2. Problem appearance
All starrocks processes are deployed and started on the same host that does not support the avx2 instruction set. It seems that FE is started, but BE is abnormal, but according to the official statement, both FE and BE services should not start.
1. Start-up phenomenon of starrocks_be:
By starting the command:
sh start_be.sh --daemon
After starting, it is found that the process does not exist:
ps -ef | grep -v "grep " | grep "/be/lib/starrocks_be"
There are only be.out files under the be/log path, and there are no log files such as be.log and be.INFO. And there is only the following information in the log:
start time: Thu Nov 24 09:59:30 UTC 2022
start time: Thu Nov 24 10:01:17 UTC 2022
start time: Thu Nov 24 10:03:15 UTC 2022
start time: Thu Nov 24 10:05:27 UTC 2022
start time: Thu Nov 24 10:08:01 UTC 2022
start time: Thu Nov 24 10:11:17 UTC 2022
start time: Thu Nov 24 10:15:50 UTC 2022
start time: Thu Nov 24 10:22:47 UTC 2022
2. Startup phenomenon of starrocks_fe
Since starrocks_fe started it has been
get bad heartbeat response: type: BACKEND, status: BAD,
msg: java.net.ConnectException: 拒绝连接 (Connection refused)
This is easy to understand, because be has not started, fe has been unable to receive be's heartbeat, resulting in connection rejection.
If fe starts up but reports such an error again, then you can determine the network problem of the communication between be and fe, and check whether the port is connected.
So far, the problem can be simply described as: because the be startup is abnormal, the activities after fe startup are also abnormal, and the root cause of the be startup exception is:
Reference:
https://docs.starrocks.io/zh-cn/latest/deployment/deployment_prerequisites
https://forum.mirrorship.cn/t/topic/4655