Forwarded from our auto-evict dev: In the case you described, if node1 and node2 had a normal replica whereas > node3 had a diskless one, the replacing of the resource will turn the > diskless replica into a diskful one. > When node1 comes back online, it will still be EVICTED. Additionally, > linstor will then actually delete the previously moved DRBD-resources. > However, everything else (such as storage pools, net-interfaces, etc.) is > still there. The node can be restored (that is, moved out of the EVICTED > status so that it can be used again) by using linstor node restore > [nodename] >
Cheers, Rene On Tue, Nov 10, 2020 at 9:30 PM Yannis Milios <yannis.mil...@gmail.com> wrote: > Hello, > > Quick question, just wondering how will "auto-evict" affect a 3 node > linstor cluster with a replica number of 2? Say that node1 goes down for > more than an 1h, linstor will try to replace its drbd resources to either > node2 or node3 assuming that the redundacy level falls below 2 and there's > enough free space in the backing device on remaining nodes (will diskless > clients count in this case?). > > How will linstor respond in the case where node1 comes back online? Will > it just restore back drbd resources on it ? or it will just reject that > node from being a part of the cluster and in this case the node will have > to be rejoined to the cluster. > > Thank you, > Yannis > > On Mon, 9 Nov 2020 at 11:16, Rene Peinthor <rene.peint...@linbit.com> > wrote: > >> Hi! >> >> This release brings 2 new added features, auto-evict and configurable >> ETCD prefixes: >> >> Auto-Evict: >> If a satellite has no connection to the controller for more than an hour, >> the controller will mark that node as EVICTED and remove all its >> DRBD-resources. Should this cause the total number of replicas for those >> resources to fall below a user-set minimum, it will then try to place new >> replicas on other satellites to keep enough replicas available. >> >> ETCD-prefixes: >> You can now configure the used ETCD prefix within the linstor.toml file, >> of course this needs to be done before the first start of the controller. >> As a little drawback(cleanup) of this change, it isn't possible anymore >> to directly upgrade ETCD backed Linstor-Controller installation from a >> version prior < 1.4.3. If you have such a situation >> upgrade to 1.9.0 first and then to 1.10.0. >> >> linstor-server 1.10.0 >> -------------------- >> * Added auto-evict feature >> * ETCD prefix is now configurable (migration only works now starting >> from version 1.4.3) >> * Block IO can now be throttled also by iops >> * Fixed REST-API single snapshot filtering >> * Fixed drbd-events2 parsing race condition >> * Fixed toggle-disk doesn't work if an unrelated node is offline >> * Fixed race-condition in auto-tiebreaker >> * Fixed usage of wait for snapshot-shipping >> * REST-API version 1.5.0 >> >> https://www.linbit.com/downloads/linstor/linstor-server-1.10.0.tar.gz >> >> Linstor PPA: >> https://launchpad.net/~linbit/+archive/ubuntu/linbit-drbd9-stack >> >> Cheers, >> Rene >> _______________________________________________ >> drbd-announce mailing list >> drbd-annou...@lists.linbit.com >> https://lists.linbit.com/mailman/listinfo/drbd-announce >> > -- > Sent from Gmail Mobile > _______________________________________________ > Star us on GITHUB: https://github.com/LINBIT > drbd-user mailing list > drbd-user@lists.linbit.com > https://lists.linbit.com/mailman/listinfo/drbd-user >
_______________________________________________ Star us on GITHUB: https://github.com/LINBIT drbd-user mailing list drbd-user@lists.linbit.com https://lists.linbit.com/mailman/listinfo/drbd-user