[Rdo-list] Swift generating thousands of rsyncs per second on a small cluster

[Diogo Vieira]

Hello,

I have a small cluster consisting on a physical machine running a Proxy and a Storage Node and 4 virtual machines running one Storage node each. Each Storage Node has only one device and the cluster has 5 zones and 3 replicas all configured with packstack.

Between the real machine (Proxy Node) and the virtual machines I have a firewall that logs all the traffic. I ran out of disk space in the firewall and came to the conclusion that the problem was the traffic generated between the Proxy Nodes and the virtual machines in port 6000 (which is of the object-server if I'm not mistaken). The problem is that the traffic being generated is on the order of a thousand or 2 rsyncs per second which seems a bit excessive. Is this behaviour normal? How much traffic should I be getting with this setup, with several (probably thousands) very small objects and less than 10 containers since it is not being accessed at all (right now we're not using the cluster)? Can someone help me understand where the problem is?

Assuming this is a problem, I tried to lower the concurrency on the object/container/account replicators and the number of workers in the proxy-server without any success.

Thank you in advance,

Diogo Vieira <dfv at eurotux.com>
Programador 
Eurotux Informática, S.A. | www.eurotux.com
(t) +351 253 680 300

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rdoproject.org/pipermail/dev/attachments/20140627/19a4497f/attachment.html>