Even we are facing a similar issue when a user or an application runs a huge query. Leader alpha service is getting restarted very frequently. When we checked our error log we found the below error messages before the service failure
Dec 13 11:43:30 dgraph: E1213 11:43:30.390352 x.go:354] Error while writing: write tcp 192.168.x.x:xxxx->192.168.x.x:xx965: write: broken pipe
Can anyone help to fix this issue?