simple-quota

It provides feature of managing blocks used below a given namespace. (Expects the inodes to be linked to the proper namespace). It follows a simple logic for quota implementation: I manage only the local data, but if someone 'trusted' tells me, there is more usage in other places, I will consider it. I don't preserve the information about cluster view, so, everytime I am restarted, someone should give me global data to be proper.

This translator is designed to be sitting on brick side (ie, somewhere above storage/posix in same graph).

More on the reasons for this translator is given in this RFC

Usage

Set Quota

One need to mark a directory as namespace first, and then set Quota on top of it.

setfattr -n trusted.glusterfs.namespace -v true ${mountpoint}/directory

To set 100MiB quota limit:

setfattr -n trusted.gfs.squota.limit -v 100000000 ${mountpoint}/directory

Updating Quota

Updating the hard limit is as simple as calling above xattr again.

setfattr -n trusted.gfs.squota.limit -v 500000000 ${mountpoint}/directory

Deleting the Quota

Call removexattr() on the directory, with the flag.

setfattr -x trusted.gfs.squota.limit ${mountpoint}/directory

Helper process

This quota feature is not complete without a global view helper process, which sees the complete data from all bricks.

Idea is to run it at regular frequency on all the directories which has Quota limit set

...
qdir=${mountpoint}/directory

used_size=$(df --block-size=1 --output=used $qdir | tail -n1);
setfattr -n glusterfs.quota.total-usage -v ${used_size} $qdir;

...

With this, the total usage would be updated in the translator and the new value would be considered for quota checks.