On stock Hadoop, I have to restart the datanode after replacing a disk. There is a JIRA on this. Is this true for MapR as well?
I have hot-swappable disks.
asked 22 Jun '11, 08:50
Disk failure detection and rectification is very easy to handle in MapR. The MapR Heatmap allows a cluster administrator to easily detect failed disks. Additionally one can subscribe to MapR alarms for disk failure.
Once any failed disks have been replaced, they can be reintroduced into the running cluster without restarting any storage servers.