Inode Cache - file operations suspend heavily creating massive latency
Applies to
- ONTAP 9
- NFS
- SMB
Issue
Overall very poor performance for file/directory based operations, with the following possible symptoms:
- Many long running READDIR/READDIR+ requests could be identified from a packet trace, it could take up to tens of seconds, or even hundreds of seconds in the worst case scenario
- Note: The same applies to SMB2 QUERY_DIRECTORY requests or its equivalents in other SMB versions
- High latency for other file operation types, such as GETATTR, LOOKUP, ACCESS, etc.
- Timeout on the NFS/SMB clients
- Extremely high latency reported from various Performance Monitoring tools
- The following
wafl.readdir.expired
error message from EMS could indicate the presence of a similar issue
[<node_name>: wafl_exemptxx: wafl.readdir.expired:error]: A READDIR file operation has expired for the directory associated with volume <volume_name>/@vserver:<vserver_uuid> Snapshot copy ID xx and inode <inode_number>.