NFS performance: How to track down a bad UNIX or Linux host
Applies to
- Data ONTAP 8.2 7-Mode
- Data ONTAP 8.1 7-Mode
- Data ONTAP 7 and earlier
Description
In NFS environments, it is possible to see sluggish performance for a single client, or subset of clients. This article discusses the different methods for troubleshooting client NFS performance for 7-mode and Clustered Data ONTAP.
7-mode:
The command ‘nfsstat’ can be used to obtain statistics regarding NFS clients. This requires per-client statistics to be enabled, which entails some performance overhead. Therefore, these statistics are not as beneficial when the performance impact experienced is wide-reaching, as it could make the problem worse. For a smaller number of clients, the overhead is usually insignificant.
To enable per-client statistics:
> options nfs.per_client_stats.enable on
Clustered Data ONTAP:
Clustered Data ONTAP does not provide per-client NFS statistics. As of ONTAP 9, the ‘statistics top’ family of commands exists to provide a list of top clients, or top files. Therefore, troubleshooting a poor performing client with Clustered Data ONTAP requires a bit of finesse.