Home > Storage > Unity XT > Storage Admin > Dell Unity: NAS Capabilities > File top talkers
Starting with OE version 5.1, file top talkers are available. This feature allows administrators to capture real-time I/O activity statistics on Dell Unity file systems and NFS datastores. This can be used to monitor the storage environment and troubleshoot performance issues. Some examples include seeing the busiest VMs on a NFS datastore or tracking down noisy neighbors or runaway clients.
Correlated statistics for SMB, NFS, files, and quota trees can be viewed using the svc_topstats service command. Each instance of the svc_topstats -monitor command displays metrics for the specified SP and statistics path (statpath). If you want to view statistics on both SPs or multiple statpaths at the same time, multiple instances of the svc_topstats can be run simultaneously. The table below shows a list of the available statpaths, a description, and the metrics that are displayed.
Statpath name | Output (per SP) | Metrics |
cifs.nassvr.*.client | Clients based on SMB activity on each NAS server | SMB operations, bandwidth, and latency |
cifs.nassvr.*.user | Users based on SMB activity on each NAS server | SMB operations, bandwidth, and latency |
cifs.nassvr.*.server | SMB servers based on SMB activity | SMB operations, bandwidth, and latency |
nfs.nassvr.*.client | Clients based on NFS activity on each NAS server | NFS operations, bandwidth, and latency |
nfs.nassvr.*.user | Users based on NFS activity on each NAS server | NFS operations, bandwidth, and latency |
nfs.nassvr.*.group | Groups based on NFS activity on each NAS server | NFS operations, bandwidth, and latency |
nfs.nassvr.*.export | NFS exports based on NFS activity on each NAS server | NFS operations, bandwidth, and latency |
nfs.filesystem.*.client | Clients based on NFS operations on each file system | NFS operations and latency |
fs.qtree | Quota trees based on total activity | Total operations and bandwidth |
fs.filesystem.*.file | Files based on total activity on each file system | Total bandwidth and latency |
When displaying metrics using the svc_topstats -monitor command, there are also optional switches that can be used to customize the output. A list of these switches, their default setting, and a description of each is displayed in the table below.
-monitor Switches (Optional) | Default | Description |
-sort <field_name> | None | Sort output by a specific field |
-order <asc | dsc> | Varies | Order output by ascending or descending values. If the sort field is a numeric value, the default is descending. Otherwise, it is ascending |
-lines <lines_of_output> | 256 | Specify the number of entries (for example, top 3 clients) in the output |
-count <count> | Unlimited | Specify the number of times to display the output before terminating |
-interval <seconds> | 15 | Specify the output interval |
-terminationsummary <no | yes | only> | Yes | Displays a summary report with the minimum, maximum, and average values of each metric |
-noresolve | Disabled | Do not attempt name resolution. IDs are displayed instead |
Here is an example of a svc_topstats command:
svc_topstats SVDM_A -m cifs.nassvr.*.client -i 1 -c 5 -l 1 -sort AvgTime -order desc -terminationsummary yes
This command displays the following statistics:
Having the various statpaths and optional modifiers available makes this feature flexible and customizable. These options can be modified as needed to fit your specific requirements. The figure below shows the output for the example command.
For more information about the svc_topstats command, reference the Service Commands document on Dell Technologies Info Hub.
When viewing statistics outputs, Dell Unity resolves IDs to names by default. Since name resolution may have dependencies on external services, the initial output may display IDs while the name is being resolved. For example, when viewing the top clients, the initial output may display IP addresses while DNS is queried for the hostname. Once the hostname is resolved, the subsequent lines automatically switch to displaying hostnames. If a name cannot be resolved, IDs continue to be displayed.
Services such as DNS, LDAP, NIS, internal system mappings, and the file resolver cache are used for name resolution. These are used to resolve items such as IPs to hostnames, UIDs/GIDs to usernames, file system IDs to file system names, quota tree IDs to quota tree paths, and inodes to filenames.
If name resolution is not needed, the -noresolve switch can be used. This option does not attempt to resolve names and only displays IDs. An example where user IDs are displayed instead of usernames is displayed in the figure below.
Inode to filename resolution is performed by the file resolver service. The file resolver service crawls specified paths, resolves inodes to filenames, and caches this information. The svc_topstats command queries the file resolver cache whenever it needs to resolve inodes to filenames.
The file resolver service is disabled by default. Caching everything by default could result in a very large cache since file systems can contain many files and directories. Due to this, svc_topstats only displays inode numbers by default. The inode numbers can be manually resolved to filenames using commands on the client.
If array-based filename resolution is needed, the administrator must enable the file resolve service and specify the paths to be crawled. The file resolver service can be managed by using the svc_topstats -fileresolver command. When starting the service, the -maxlimit switch can be used to specify the maximum number of entries in the cache. The acceptable range is 100K – 2 million and the default setting is 1 million. Once the limit is reached, no additional entries are added to the cache unless some are dropped first. This limit can be changed at any time but requires a restart of the file resolver service to take effect. The figure below shows an example of checking the service status, starting the service with a 2 million entry limit, and stopping the service.
After the service is started, the administrator also needs to specify which paths the service should crawl. The path should be in the format /<NAS_Server>/<File_System>/<path> such as /NAS1/FS1/Dir1/. Paths can be dropped from the crawler’s list if they are no longer needed. If a path is dropped, its cached entries remain available until the service is restarted. It is recommended to remove unused paths and stop the service when inode to filename resolution is not needed. The figure below shows an example of displaying the list of paths to be crawled, adding a new path, and dropping an existing path.