Performance.cache-size and server node RAM usage

[edrock200]

Sorry if this is a dumb question, but I can't seem to find the answer in the documentation and google results yield different answers. Is the value of performance-cache.size the amount of ram used on the gluster server nodes, or on the clients? I can set it really high if servers, but not nearly as high if on clients. Thanks in advance for any insight.

Second question, which may seem odd, but my gluster nodes are dedicated and have a good chunk of RAM, 128GB. Presently only ~8% if used max per node. Is there a way to get gluster to use more RAM for caching and such?

[edrock200]

Well, in case anyone else has this question, it's definitely client ram. As a test I cranked it up to more than my test VM client had in ram, and that VM wouldn't mount with a cache ram needed error in the gluster log. Lowered it to less than the vms ram, and it mounted. Interestingly this setting also doesn't update on the fly. It appears you must remount clients after changing this setting. Is there a way to make suggested documentation updates for the gluster manual?

As I mentioned above, unless I missed it, the performance cache setting documentation does not make it clear where and how this setting will be used. In. Googling it, you will find several blog posts in which the posters believe this setting uses ram on the gluster nodes vs client end points, further adding to the confusion. But given this setting uses client RAM, this could lead to bad things if a sysadmin doesn't realize he will be chewing up that amount of ram on every client endpoint.