We have some AFS volumes which hold data for various web servers on campus. These volumes are replicated on multiple file servers, because that's an intelligent thing to do with important data.
We also have multiple web servers, because it's also intelligent to split the web-serving load across machines, and to have redundancy in our web servers. All of these are virtual machines on various VMware hypervisors. While doing some benchmarks on new AFS file servers, I noticed that AFS data is transferred much faster if the afs client machine is on the same hypervisor as the afs file server that it's contacting. Is there some way to encourage this to happen on our web servers? If the important AFS data is replicated on both afsfs1 (on vmhyper1) and afsfs2 (on vmhyper2), can we have a web-server on vmhyper1 prefer afsfs1 instead of afsfs2? I'd want it to still use afsfs2 if it can't reach afsfs1 (for the case where vmhyper1 is fine, but afsfs1 is being rebooted for some reason). Nothing urgent here. Just wondering. -- Garance Alistair Drosehn = [email protected] Senior Systems Programmer or [email protected] Rensselaer Polytechnic Institute; Troy, NY; USA _______________________________________________ OpenAFS-info mailing list [email protected] https://lists.openafs.org/mailman/listinfo/openafs-info
