Ask Your Question

Revision history [back]

Thanks for providing so much feedback, this sort of open kimono down in the weeds turning from a variety of deployments is exactly what's needed to understand true "best practices" rather than a collection of cargo cult information silos.

The storage nodes are combined account/container/object yes? What is your part power and how much RAM and how many devices do you have in your storage nodes? Do you see any signs of swap usage when the load is high? You might try cranking up vfs_cache_pressure.

When you created the filesystem what did you specify for the i node size (you can check "isize" with sudo xfs_info /dev/sdX)?

Thanks again for sharing your experience.