[OpenAFS] vos release fails -- bandwidth too low?
Sun, 31 Aug 2008 16:12:12 -0800
We have a multi-site setup with OpenVPN and are evaluating technologies =
for serving files to local users and across the vpn. From what I've read, =
OpenAFS is a good match for our low-bandwidth vpn links, but in the actual =
implementation it isn't working out.
We've implemented a bridged OpenVPN multi-site network via UDP. The vpn =
has not been specifically optimized for bandwidth, but that is probably a =
discussion for another forum. We are currently seeing about 26KBps =
sustained across the links.
Our OpenAFS trial setup has two file servers, one with an RW volume and =
both with RO volumes of the RW one.
In our simple tests, we made a change to the RW volume and attempted a =
release. In the case where both servers are at the same site (and =
therefore have the full lan bandwidth), a vos release works perfectly. In =
the case where the two servers are at different sites, the vos release =
fails for all practical purposes: that is, it never finishes and there =
doesn't seem to be any serious amount of communication across the vpn =
between the two servers. There are no relevant error messages in the afs =
logs, except for the every 30-second info/warning messages about the vos =
release continuing. There are obviously firewalls involved when operating =
across the vpn, but it is a fully-bridged network and an analysis of =
dropped packets at the firewalls and the file server machines does not =
show anything related to AFS.
Where should we go from here? Is our vpn bandwidth too low to support =
openafs? Is there some additional analysis to perform? Is there a better =
afs setup to use? Any hints on vpn tweaks to resolve the issue?
Thanks in advance,