VFS Stats
Overview: Collection of the "vfsstats" output from all nodes running gpfsmond.
Graph Descriptions
read / write - graph per cluster, combined across all clusters
| Item | Description |
|---|---|
| Horizontal Axis | time stamp |
| Vertical Axis | calls per second |
| reads | on positive axis |
| writes | on negative axis |
metadata - graph per cluster, combined across all clusters
| Item | Description |
|---|---|
| Horizontal Axis | time stamp |
| Vertical Axis | calls per second |
| access | opens/closes/etc. |
| metadata | getattr/setattr/create/delete/etc. |
| directory | mkdir/rmdir/readdir/etc. |
filesystem / nfs - graph per cluster, combined across all clusters
| Item | Description |
|---|---|
| Horizontal Axis | time stamp |
| Vertical Axis | calls per second |
| filesys | statfs/etc. |
| nfs | NFS-related calls |
est. average request size - combined across all clusters
| Item | Description |
|---|---|
| Horizontal Axis | time stamp |
| Vertical Axis | Bytes |
| reads | on positive axis |
| writes | on negative axis |
Data Files
Data Source
This data is collected from gpfsmond collecting the vfsstats output from GPFS.
Ganglia Processing
All vfsstats data is divided by 15 in order to show calls/second instead of total calls.
The "Est. average request size" graph divides the estimated yellowstone/DAV GPFS bandwidth by the number of yellowstone/DAV read/write calls.