Cluster deployment, More than 100 concurrent requests, two exceptions:
- Abnormal connection
2.Alpha node is down, resulting in the failure of the connection request from the service to the node.
The connection fails even after the node is restarted.
Deployment information: 1 zero, 2 alpha groups (6 alpha nodes in total), and 1 million nodes in the library.
。
One request triggers two queries:
I’m sorry, what exactly is the issue? The first issue - abnormal connection - I believe Dgraph should be able to automatically recover… As for issue 2, could you elaborate more on why the alpha node is down?
Yes, the first problem does recover automatically, but is there any way to avoid it, or why？
There should be a large number of concurrent requests. I tried 100, 300 and 500 concurrent requests, and all of them had node downtime.
I set the log level to 3. The logs of the two down nodes are not the same, and the zero node has no exception logs. I want to know if it is not supported at present, but I don’t have this problem when I use standalone.
Sorry, my description may not be clear. First of all, I configure the dgraph cluster directly on a machine and query through the Java client. Secondly, Java clients (not distributed) are deployed on another machine, and 100 to 500 concurrent requests are initiated through JMeter. Finally, the above problems are caused.