[DownstreamMerge] Merge 3-4-22 by trozet · Pull Request #987 · openshift/ovn-kubernetes

trozet · 2022-03-04T20:28:13Z

@martinkennelly @tssurya @dcbw @msherif1234 FYI

…init In order to avoid any potential bringup race conditions between dbchecker container and sbdb/nbdb we move the memory trimming check and enable to start of each db. Signed-off-by: Mohamed Mahmoud <mmahmoud@redhat.com>

We need to stop the node IP manager before the factory, becuase the node IP manager uses the factory. Fixes: eb73081 node: wait for nodeIP and OpenFlow manager to stop before next test Signed-off-by: Dan Williams <dcbw@redhat.com>

A metric is the first step towards telemetry and we need to know the count of lgw users. This will be a first step towards that. Signed-off-by: Surya Seetharaman <suryaseetharaman.9@gmail.com>

gateway: fix test teardown order to really prevent nodeIP manager flake

We are using labelled (vector) metrics to export info and if the label values change, we still export the old labels values which confuses users because they have no way of determining which labelled metric is the latest. We must reset each vector metric prior to using it to ensure we are only exporting the current state and not old states which should have been scraped previously due to our usage of labels that can be dynamically updated during runtime. This is occurring in the OVN metrics exported by ovnkube-node. Follow on work is needed to fix OVS metrics that are used by ovs-exporter. Signed-off-by: Martin Kennelly <mkennell@redhat.com>

Add metric to record gateway modes

* CI: Retest only failed E2E jobs Only retest failed E2E jobs out of the matrix. Do the same for DualStack and Upgrade tests. Test results will be stored inside github's cache and if a sub-job was marked as completed, it is skipped and its results from the run before are going to be displayed instead. Also cache master and PR builds. Signed-off-by: Andreas Karis <ak.karis@gmail.com>

It's all the same code, move some things around so they all look the same in preparation for the next step. Signed-off-by: Dan Williams <dcbw@redhat.com>

Signed-off-by: Dan Williams <dcbw@redhat.com>

Move memory-trimming-on-compaction out of dbchecker to nbdb/sbdb init

- To run the setup script from a container - To explicitly supply a kubeconfig - To explicitly name the cluster - To provide a custom service DNSdomain - To itegrate a locally running Docker Registry Remove bits that manually configure apiserver-ip for the kind cluster since it could prove to be a security venerability in CI. Instead always default to the node's local loopback address as suggested in KIND documentation. Signed-off-by: astoycos <astoycos@redhat.com>

Centos is now EOL so it no longer makes sense to build ovn-k images for it. This commits deletes the centos based dockerfile and removes the associated make images step. Signed-off-by: astoycos <astoycos@redhat.com>

Remove this dockerfile because it is not needed to build ubuntu based arm images, it is a direct copy of Dockerfile.ubuntu, and it is not used within the codebase anywhere. Signed-off-by: astoycos <astoycos@redhat.com>

Automatically push the built ovn-k images with fedora, and ubuntu bases(arm64 amd6d) to ghcr.io for the ovn-org. New images will be built and push every time new changes are merged into the ovn-kubernetes master branch. Update the ovn-docker-images workflow to only run when code is merged into master. Signed-off-by: astoycos <astoycos@redhat.com>

Update the location of the KUBECONFIG file from admin.conf -> ovn.conf Signed-off-by: astoycos <astoycos@redhat.com>

egressgw cleanups and code consolidation

Upstream Submariner kind hacks, setup automatic build and storage of ovn-k images

We should only register and start the metrics if we are current leader otherwise we are confusing the user with 0 value metrics on non-leaders and wasting CPU cycles. Signed-off-by: Martin Kennelly <mkennell@redhat.com>

Prior to this change, timestamping of NB and IPSec both registered and started their metrics when ovn kube master was not leader. Functionality stays the same except 'e2e_timestamp' metric is registered with the non-ovn registry because this registry is not exposed in ovnkube master. Signed-off-by: Martin Kennelly <mkennell@redhat.com>

We do not want to expose the majority of our metrics when we are not a leader. This PR registers most of the metrics only when leader. The metrics registration are grouped according to their task to aid code readability only. Signed-off-by: Martin Kennelly <mkennell@redhat.com>

9598c132c2616 Fixes client error handling fc5e884e9f41a Moves client error handling from the cache to client 87acccaf29896 client: use LastTransactionID when the only monitor is a CondSince one 5b14611220b61 test: Using wait api with integration bb1929ac6d2ec Mapper: FieldByColumn error message shall contain table name ebe15cf84d8c4 Fix libovsdb server close 976fa4167e523 adding Wait to client api 6791b3ba1f1ce api: support pointer to slice of Models for List Signed-off-by: Dan Williams <dcbw@redhat.com>

Signed-off-by: Dan Williams <dcbw@redhat.com>

In some scale tests we get client-side throttling. Signed-off-by: Dan Williams <dcbw@redhat.com>

vendor: bump libovsdb to c8b4494412b1a0ba1396dd162a9a2d497c80f2b0

Previous to this change, we spawned a go routine without checking if the event is relevant. This will lead to excessive go routine creation. Follow on work is needed to add a queue. Signed-off-by: Martin Kennelly <mkennell@redhat.com>

Signed-off-by: Tim Rozet <trozet@redhat.com>

kube: bump kube API QPS to 50

Signed-off-by: Surya Seetharaman <suryaseetharaman.9@gmail.com>

…etrics Metrics: Remove stale label metric entries

openshift-bot · 2022-03-05T17:39:09Z

/retest-required