I was looking at upgrade procedure for gluster I've seen that the raccomenfedd procedure is with downtime, shutting down all clients and after that upgrade gluster Is that true? Very strange, on huge clusters with hundreds of clients and thousands of virtual machines (gluster is a scale-out storage so huge clusters should be normal) would be a mess and a lot of time and money loss I'm not using gluster but i know for sure that i can't shutdown all of my VMs (actually about 190) for an upgrade This means shutting down all redundancy, in example I'll have to shut down all nameservers. All of my domains will stop working even the ones pointing outside my network No DNS =no mails, no sites and so on -------------- next part -------------- An HTML attachment was scrubbed... URL: <http://www.gluster.org/pipermail/gluster-users/attachments/20160526/bcc6ca3a/attachment.html>
On 05/26/2016 11:57 AM, Gandalf Corvotempesta wrote:> I've seen that the raccomenfedd procedure is with downtime, shutting > down all clients and after that upgrade gluster >Our upgrade procedure is to upgrade the servers first (shutdown the Gluster service on a server, upgrade that server, then reboot, then go to the next server once it has come back online and sync'ed), then the clients one by one. No downtime, except rebooting the individual machines one by one as running tasks allow. Gluster services our HPC Grid so it is not uncommon to have a client on the previous version for several days until jobs finish running on it so it can be rebooted. Note this is going between minor revisions, like 3.7.x to 3.7.y. -Dj