Thanks Andreas! It totally makes sense that it would be an oppressive task to test every single client version against each new server version that comes out. I guess I was hoping for some real-world experience (either “worked for us” or “we had problems”) where someone had used an older 2.5 Lustre client to access the newer 2.10 Lustre server.
Mark -- Mark Miller – JHPCE Cluster Technology Manager Johns Hopkins Bloomberg School of Public Health Office E2530, 615 N. Wolfe St., Baltimore, MD, 21205 443-287-2774 | https://jhpce.jhu.edu/ From: Andreas Dilger <[email protected]> Date: Tuesday, July 3, 2018 at 11:25 PM To: Mark Miller <[email protected]> Cc: "[email protected]" <[email protected]> Subject: Re: [lustre-discuss] Lustre 2.10.4 server with 2.5.32 clients Lustre has a fairly robust network protocol negotiation at mount time between the client and server, and we try hard to avoid any protocol changes that are not covered by the connect-time negotiation. However, we aren't able to test all possible combinations of versions, so the published list is what we have tested. It is entirely possible that older client versions will work with newer server version, and vice versa. Cheers, Andreas On Jul 3, 2018, at 10:57, Mark Miller <[email protected]<mailto:[email protected]>> wrote: Hi All, A while back, someone asked the list about using Lustre 2.10.4 clients to access a Lustre 2.5.3 server, and the consensus was that it worked without issue. We have the opposite scenario where we have RHEL6.9 compute nodes running version 2.5.32 of the Lustre client, and we are standing up a new Lustre server running version 2.10.4. It seems to work with the testing I’ve done, but I wanted to see if anyone has run across any issues with running older Lustre clients against newer Lustre servers (and also see if anyone has run that combo successfully). Part of the concern comes from http://wiki.lustre.org/Lustre_2.10.4_Changelog which says lists “Interoperability Support” as “Clients & Servers: Latest 2.9.X” We are planning on upgrading the Lustre clients on the compute nodes during the next patching cycle, but upgrading now would require quiescing the nodes to unmount the lustre filesystems and unload the lustre module. Thanks for any insights!! Mark -- Mark Miller – JHPCE Cluster Technology Manager Johns Hopkins Bloomberg School of Public Health Office E2530, 615 N. Wolfe St., Baltimore, MD, 21205 443-287-2774 | https://jhpce.jhu.edu/ _______________________________________________ lustre-discuss mailing list [email protected]<mailto:[email protected]> http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
_______________________________________________ lustre-discuss mailing list [email protected] http://lists.lustre.org/listinfo.cgi/lustre-discuss-lustre.org
