Releases: threefoldtech/zos
v0.4.0-rc9
Merge pull request #925 from threefoldtech/dev/check-hw-addr WIP: clean up network on container destroy
v0.4.0-rc8
Merge pull request #917 from threefoldtech/dev/zdb-improvement Handle zdb namespace and macvlan cleanly
v0.4.0-rc7
- Hot fix #921
v0.4.0-rc6
fix broken counters provisiond, closes #874
v0.4.0-rc5
Merge pull request #878 from threefoldtech/networkd_migration Networkd migration
v0.4.0-rc4
limit the ygg node name to 6 characters this is mainly esthetic and avoid creating garbage on the network map: http://[21f:dd73:7cdb:773b:a924:7ec0:800b:221e]
v0.3.5
6aa6a8b
zaibon
Christophe de Carvalho
Changelog
v0.3.5 mostly focuses on networking. In this version, 0-OS introduces a new network daemon which allows nodes without public IPv6 address to be able join the grid and be reachable by all other nodes.
To accomplish this, 0-OS have integrated Yggdrasil, a network daemon creating a fully end-to-end encrypted IPv6 overlay network and employs routing algorithm based on a globally-agreed spanning tree.
You can find more information about Yggdrasil at: https://yggdrasil-network.github.io/
With yggdrasil in place, nodes that were previously incapable of running 0-OS v2 because they only had access to IPv4 network can now upgrade and join the grid.
A second side effect of this, is the way how nodes report uptime to the grid. Previously, any booted nodes would send its uptime report to the explorer. After this release only nodes that are full ready to process reservations will report uptime. This should greatly simplify the way nodes are chosen by user and avoid creating reservation that are never picked up by the nodes.
Some more documentation around yggdrasil integration can be found at: https://github.com/threefoldtech/zos/blob/master/docs/network/yggdrasil.md.
This version also contains some version upgrade for
0-OS main daemons
capacityd
- re-enable capacityd blocking on networkd
after this commit, capacity won't report node uptime anymore unless networkd is fully booted
contd
- Support WithBinaryLogURI from containerd to restart logs shim on kill
- Update shim-logs to v0.2
networkd
- #825 simplify the routing table of the network resources
leaves wireguard deals with routing and only set a single route for the full network range - do remove address from wg interface if needed
- allow public namespace to also receives IPV6 configuration from router advertisement
- enable accept_ra and accept_ra_defrtr in public namespace
- allow to have an yggdrasil interface into 0-db containers
- make ndmz an interface and implement ipv4 only and dualstackversion
- disable forwarding on zos bridge. this to avoid losing default route if router advertisement has been restarted
- choose only ipv4 yggdrasil peer when running in ipv4 only mode
- close zinit connection after yggdrasil is stopped
this fix a bug where the yggsrasil service wasn't properly stopped cause the zinit client was closed before the stop signal was sent - ensure we always skip yggdrasil public peer from the same segment
this is to ensure that the node doesn't only connect to other node from the same farm and have at least on other public peer connected to yggdrasil network - add a fallback list of public yggdrasil peer to be used in case we cannot reach the online version of the public peer list
provisiond
- only allow to create container root filesystem on top of SSD
- implement zdb upgrade procedure
- #839 fix tracking of resource unit consumed by 0-DB namespaces
- #850 make corex only listen on the private overlay network interface of the container. this allow container with public IPv6 to NOT expose coreX
extra binaries
- Add 0-fs to the extra bins workflow
- Add coreX to the extra bins workflow
- Update corex to 2.1.3 (musl issue fixed)
- Add yggdrasil to the extra bins workflow
containerd
Update containerd to v1.14
v0.4.0-rc3
Update CODEOWNERS
v0.4.0-rc2
Merge pull request #853 from threefoldtech/fix_tests avoid skipping failed tests
v0.4.0-rc1
go mod tidy