New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Upstream merge 2020-04-11 #134
Upstream merge 2020-04-11 #134
Conversation
Pass MACs around as net.HardwareAddr internally, and only convert to strings when we actually want them as strings. (This is simpler than the alternative because "mac.String()" can't fail, but "net.ParseMAC(macStr)" can.) Signed-off-by: Dan Winship <danw@redhat.com>
As with MAC addresses, pass IPs/CIDRs around internally as the parsed types and only convert to string when needed. Signed-off-by: Dan Winship <danw@redhat.com>
Signed-off-by: Andrew Sun <asun@redhat.com>
Changes-Include: - Use verbose print resource usage - Use vendor folder to avoid having to download deps - Bump Timeout to 15 min Using vendor for download saves about a minute; we see that the entire execution is around 9-10 minutes, while our timeout was 10 min. Closes openshift#1222 Signed-off-by: Tim Rozet <trozet@redhat.com>
Fix golang linter timeouts
This commit: 1. Removes the template YAML file and use of Ruby to handle YAML anchors 2. Squahes the Verify and Build Job as they have identical node requirements 3. Uses a matrix for the sharded KIND e2e jobs 4. Moves the Golang Version to an env variable to avoid duplication 5. Adds a Makefile to the test directory CI can then simply `make -C test $TARGET` to run tests. This is preferable to having the logic inside the CI workflow as it can be used for local testing Signed-off-by: Dave Tucker <dave@dtucker.co.uk>
This commit adds the HA testing back in to the matrix. It was easier to add this in after the ruby was removed than to rebase. Signed-off-by: Dave Tucker <dave@dtucker.co.uk>
Run using: OVN_HYBRID_OVERLAY_ENABLE=true OVN_HYBRID_OVERLAY_NET_CIDR=<overlay_prefix> pushd contrib/; ./kind.sh Signed-off-by: Brent Salisbury <bsalisbu@redhat.com>
Signed-off-by: Andrew Sun <asun@redhat.com>
ci: Remove ruby dependency and simplify workflow
pass MACs, IPs, CIDRs around in parsed form rather than as strings
Add e2e test that kills ovnkube-master; fix ovnkube-node test for multiple nodes
In Ovnkube.sh, currently Node IP's are obtained using DNS. Instead, Pod spec (status.HOST IP) is used to obtain the Node IP's. Signed-off-by: Pardhakeswar Pacha <ppacha@nvidia.com>
Currently, ovn_log_northd variable captures the loglevel which should capture the log location. Renaming ovn_log_northd -> ovn_loglevel_northd and similar renaming is used for other variable names (ovn_log_nb, ovn_log_sb, ovn_log_controller). Similarly in the YAML files, OVN_LOG_NB and other log level variables(OVN_LOG_SB, OVN_LOG_NBCTLD) is used for for capturing the loglevel. Renaming them to OVN_LOGLEVEL_NB and others to similar nomenclature. Signed-off-by: Pardhakeswar Pacha <ppacha@nvidia.com>
Signed-off-by: Dan Winship <danw@redhat.com>
The CNI code had code to check if there was a default route specified in the "routes" field of the pod annotation and ignore the "gateway_ip" field in that case, but we never actually used that functionality. Remove the code, and sanity-check the routes when marshalling/unmarshalling the pod annotation. Signed-off-by: Dan Winship <danw@redhat.com>
Signed-off-by: Dan Winship <danw@redhat.com>
- currently the variable is named with _test suffix, as in, ovn_nbdb_test. change it to ovn_nbdb_conn - add similar variable for SB DB - use the above two variables everywhere in the script Signed-off-by: Girish Moodalbail <gmoodalbail@nvidia.com>
Signed-off-by: Pardhakeswar Pacha <ppacha@nvidia.com>
Signed-off-by: Andrew Sun <asun@redhat.com>
Signed-off-by: Andrew Sun <asun@redhat.com>
nit: change ovsdb server connection method variable name
ovnkube.sh: Adding iptable rules to open up the OVN NB/SB DB raft ports
Download kind in CI instead of go getting it
Add hybrid-sdn support to KIND
update pod annotation and CNI shim communication for dual-stack
Instead of "End-To-End Tests (e2e-k..." in the Github show more of the relevant bits eg "e2e (shard-n,...". Since all the tests use KIND, and all the tests use OVN, "kind-ovn-" seems less relevant too. Signed-off-by: Dan Williams <dcbw@redhat.com>
ci: shorter GitHub action job names
Also run CI on non-master branches
In particular, the loadbalancer functions are used for all types of services (ClusterIP, NodePort, ExternalIP) so it's confusing to refer to "service IP" when the IP might actually be, eg, a node IP. So say "source IP" instead. Signed-off-by: Dan Winship <danw@redhat.com>
Going forward officially supported install methods will be KIND (Kubernetes-in-Docker) and daemonset YAML files. Signed-off-by: Dan Williams <dcbw@redhat.com>
install: remove unsupported install methods
And add explicit tests for the old format (tcp://1.2.3.4:6641) Signed-off-by: Dan Williams <dcbw@redhat.com>
Signed-off-by: Dan Williams <dcbw@redhat.com>
config: switch tests/docs normal DB address format (eg tcp:1.2.3.4:6641)
Rename createLoadBalancerVIP to createLoadBalancerVIPs and make it able to handle multiple source IPs, of multiple families, and filter the endpoints to match the source IP family/families. Signed-off-by: Dan Winship <danw@redhat.com>
Kubernetes has "masters" and "nodes". "Minion" was deprecated years ago, and "master node" doesn't mean anything, unless you mean a master which is also a node, but most of the uses of "master node" here didn't necessarily mean that. Signed-off-by: Dan Winship <danw@redhat.com>
-- assumes that all the required private keys and corresponding signed certificates are mounted into the container at /ovn-cert path -- the private keys and certificates for various OVN components are named as below ovncontroller-cert.pem ovncontroller-privkey.pem ovnnb-cert.pem ovnnb-privkey.pem ovnnorthd-cert.pem ovnnorthd-privkey.pem ovnsb-cert.pem ovnsb-privkey.pem -- the name of the CA certificate that signed the CSRs is ca-cert.pem -- disabled by default since there are lot of pre-requisites to get this thing to work. to enable it, one needs to set the OVN_SSL_ENABLE environment variable in each of the container Signed-off-by: Girish Moodalbail <gmoodalbail@nvidia.com>
error messages in GatewayInit() and GatewayCleanup() functions doesn't have the gatewayRouter name that makes it diffuclt to know for which k8s node the functionality has failed. Signed-off-by: Pardhakeswar Pacha <ppacha@nvidia.com>
…logy Fix references to "minions" and "master nodes"
SSL for OVN daemonsets
…ncers initial dual-stack loadbalancer support
/lgtm |
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: dcbw, rcarrillocruz The full list of commands accepted by this bot can be found here. The pull request process is described here
Needs approval from an approver in each of these files:
Approvers can indicate their approval by writing |
@dcbw: All pull requests linked via external trackers have merged: . Bugzilla bug 1814098 has been moved to the MODIFIED state. In response to this:
Instructions for interacting with me using PR comments are available here. If you have questions or suggestions related to my behavior, please file an issue against the kubernetes/test-infra repository. |
Mostly IPv6/dual-stack updates from @danwinship
@openshift/networking