A yarn service unhealthy status processing
Since nine new nodes have been added to the hadoo cluster, it is sometimes found that the yarnweb page has one or more nodes with unhealthy status, but it returns to normal after a while, but this state often causes the alarm of background monitoring. Checked the log of the exception node
Display as follows
Yarn's local log directory was used more than 90%, and then df looked at it and found that the root directory was used at 100%.
No wonder the size of 30 gigabytes
Comparing the disk size of the new and old nodes, it is found that the size of the old nodes is 20T.
Modified the newly added node yarn configuration file to redefine the intermediate state directory, and returned to normal after restarting the service.
