0

I have installed GlusterFS in node A,B,C. The node A and B were GlusterFS servers (brick1-4), meanwhile, they were GlusterFS clients for creating files. And the node C didn't exist bricks, it was just a GlusterFS client to mount by mount -t glusterfs nodeA:/XXX /mnt/glusterfs.

Here is the question:

  1. In node C, I executed time ls /mnt/glusterfs, first time was 10s, the second was 0.3s, it got the dentry from the cache. By free -m in node C, it didn't changed at all. So I executed free -m in node A and B, the memory in used has grown.
  2. In other words, the dentry was cached in the server, the client didn't cache the dentry.
  3. How to explain this phenomenon? By the way, Gluster version is 3.3.0.

Here is the test, controller node and compute2 node are two GlusterFS servers, compute1 node is GlusterFS client(without bricks), mounted to /mnt/glusterfs.And I excuted ls.sh in compute1 node to list the directory.
the result of 'free -m before and after the ls.sh in compute1' Controller node and compute2 node have cleared the cache by 'echo 3 > /proc/sys/vm/drop_caches', the result of 'free -m' before and after the ls.sh have shown in the picture.

the result of 'free -m' before and after the ls.sh in controller node

user6481589
  • 71
  • 2
  • 6

0 Answers0