On 02/02/2018 10:26 AM, Ravishankar N wrote: >>> 2) "Replace MD5 usage to enable FIPS support" - Ravi, Amar > > + Kotresh who has done most (all to be precise) of the patches listed in > https://github.com/gluster/glusterfs/issues/230 in case he would like > to add anything. > > There is a pending work for this w.r.t rolling upgrade support. I hope > to work on this next week, but I cannot commit anything looking at other > things in my queue :(.
I have this confusion reading the issue comments, so if one of the servers is updated, and other server(s) in the replica are still old, then self heal deamon would work, without the fix? From the comment , I understand that the new node self heal deamon would crash. If the above is true and the fix is to be at MD5 till <condition> then that is a must fix before release, as there is no way to upgrade and handle heals, and hence not get into split brains later as I understand. Where am I going wrong? or, is this understanding correct? > To add more clarity, for fresh setup (clients + servers) in 4.0, > enabling FIPS works fine. But we need to handle case of old servers and > new clients and vice versa. If this can be considered a bug fix, then > here is my attempt at the release notes for this fix: > > "Previously, if gluster was run on a FIPS enabled system, it used to > crash because gluster used MD5 checksum in various places like self-heal > and geo-rep. This has been fixed by replacing MD5 with SHA256 which is > FIPS compliant." > > I'm happy to update the above text in doc/release-notes/4.0.0.md and > send it on gerrit for review. I can take care of this, no worries. I need the information, so that I do not misrepresent :) provided any which way is fine...  https://github.com/gluster/glusterfs/issues/230#issuecomment-358293386 _______________________________________________ maintainers mailing list email@example.com http://lists.gluster.org/mailman/listinfo/maintainers