On 10/12/2017 12:33 PM, Paul Heinlein wrote: > We encountered a weird problem today, and I thought some of you might > like to hear the solution. > > The underlying change was listed in the 7.4 changelog, so it's not a > bug, but it may drive you buggy. > > The majority of our HPC cluster nodes run CentOS 7, though the exact > patch levels vary from node to node. None is older than 7.3, but a few > newer nodes were kickstarted right to 7.4. > > The problem was that our mounts of Isilon NFS exports were failing > randomly among the nodes. Routing was fine. Network connectivity was fine. > > The short answer is that the default in 7.4, and I think in the > nfs-utils-1.3.0-0.48.el7 package in particular, has changed. While NFS > v4.0 was the default up to 7.3, the 7.4 protocols are subtly different: > > 1. Try NFS v4.1 first > 2. Fail down to NFS v3 > 3. Fail down to NFS v2 > > The problem is that our Isilon works with NFS v4.0, not 4.1, but 4.0 is > not in the fail-down path. > > The short-term answer is to specify nfsvers=4.0 in our autofs > configuration files, which works like a charm. > > Like I said, this was an announced change, but the implications escaped > us until now. So this little writeup is just for the record. > You are not the first person to have this issue .. thanks for the post.
Attachment:
signature.asc
Description: OpenPGP digital signature
_______________________________________________ CentOS mailing list CentOS@xxxxxxxxxx https://lists.centos.org/mailman/listinfo/centos
