gcloudnfs's Introduction

Dependencies

python gcp sdk (pip install --upgrade google-api-python-client) --> https://cloud.google.com/compute/docs/tutorials/python-guide#cloud-sdk

To create an NFS server

./gcloudnfs create --project <project> --zone <zone> --network <network> --machine-type <machine-type> --server-name <server-name> --data-disk-name <data-disk-name> --data-disk-type <data-disk-type> --data-disk-size <data-disk-size>

E.g.

./gcloudnfs create --project my_project --zone us-east1-c --network default --machine-type n1-standard-1 --server-name test-server --data-disk-name test-data-disk --data-disk-type pd-standard --data-disk-size 2000

To destroy an NFS server

./gcloudnfs destroy --project <project> --zone <zone> --network <network> --data-disk-name <data-disk-name> --server-name <server-name>

E.g.

./gcloudnfs destroy --project my_project --zone us-east1-c --network default --data-disk-name test-data-disk --server-name test-server

Reusing an existing disk

Passing the --reuse-data-disk flag to the create and destroy commands is necessary to reuse a disk.

Passing --reuse-data-disk to create causes the command to skip disk creation, instead attaching the existing disk named --data-disk-name to the existing instance.

Passing --reuse-data-disk to destroy causes the command to skip disk destruction, saving the disk for later use.

Future Performance Improvements

Add a dedicated ZIL device.
Optimize ARC cache size
Add dedicated SSDs for L2ARC cache
Add more disks and use striping with parity across the disks.
optimize async write percentages
If NFS is the bottleneck, try optimizations here : http://nfs.sourceforge.net/nfs-howto/ar01s05.html

Recommend Projects