[HDFS-3736] Failure in starting NN due to fsimage loading failure

• Type: Bug
• Status: Resolved
• Priority: Major
• Resolution: Duplicate
• Affects Version/s: 2.0.0-alpha
• Component/s: ha, namenode
• Assignee: Andrew Wang
• Reporter: suja s
• Created: 30/Jul/12 12:58
• Updated: 05/Sep/13 08:22
• Resolved: 05/Sep/13 08:22

Description

Came across a situation as follows in our test environment.
NNs running in HA mode.
While uploading checkpoint, MD5 file renaming from tmp to actual file failed due to some reason which is unknown (non IO exception).
At the same time at standby side, connection imeout occured.
This lead to tmp MD5 file and original fsimage file (ckpt fsimage file was renamed successfully to original fsimage file) in the name dir of active NN.
On NN restart it checks for MD5 file and since it is not found, startup is failing.

Issue Links

• is duplicated by: HDFS-4596 Shutting down namenode during checkpointing can lead to md5sum error
• relates to: HDFS-3277 fail over to loading a different FSImage if the first one we try to load is corrupt