[ovn-controller] Change startup mechanism of ovs pods #423
Conversation
|
Merge Failed. This change or one of its cross-repo dependencies was unable to be automatically merged with the current state of its repository. Please rebase the change and upload a new patchset. |
|
[APPROVALNOTIFIER] This PR is APPROVED This pull-request has been approved by: averdagu 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 |
|
Merge Failed. This change or one of its cross-repo dependencies was unable to be automatically merged with the current state of its repository. Please rebase the change and upload a new patchset. |
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/9d1e6fb949f14b0f902e9c4913239d6e ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 24m 47s |
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/0a45c0f113754dc8af4e86b200206bc7 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 26m 52s |
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/32fe97f422124ae882519d53895dc2c8 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 21m 05s |
averdagu
force-pushed
the
ovs-restart
branch
2 times, most recently
from
b94efec to
f304f7c
Compare
| } | ||
|
|
||
| TLSOptions="--certificate=/etc/pki/tls/certs/ovndb.crt --private-key=/etc/pki/tls/private/ovndb.key --ca-cert=/etc/pki/tls/certs/ovndbca.crt" | ||
| DBOptions="--db ssl:ovsdbserver-nb.openstack.svc.cluster.local:6641" |
There was a problem hiding this comment.
Only supporting ssl, need to add mechanism to support also non tls connections
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/83bd56facf314f1f9e13f3c122caadd9 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 22m 12s |
|
/recheck |
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/3ff505fab9994c8ab22909f47cb474a4 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 20m 16s |
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/107eac5bca244667b5651800bb406375 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 15m 47s |
averdagu
force-pushed
the
ovs-restart
branch
2 times, most recently
from
1a1a179 to
30fec14
Compare
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/9c9fa9ecc3e44aebb4d808d86e48cdf0 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 20m 03s |
This commit aims to modify the starting scripts of the ovn-controller-ovs daemonset. This is done to allow modifying the RollingUpdate Strategy to not allow any Unavailable pod during update, what this will cause is that during an update to the pod (delete old one and create new one) instead of first deleting the first one and then deleting the next one, it will create the new pod while the old one is running. Due to how the startup scritps works currently this is not allowed. The reason behind this is to try to lower the downtime observed if the environment is using centralized floating ip during an update. With this commit the ovn-controller-ovs will share the PID namespace with the host, in order to allow signaling between old/new pod. Another change is adding an STATE that the containers (ovsdb-server, ovs-vswitchd and ovsdb-server-init) will handle internally. The differents states are: - NULL (No file): will happen the fist time ds is created on the oc worker. - INIT: First time init-ovsdb-server is executed on the oc worker. - OVSDB_SERVER: once ovsdb-server pod has run the startup script - RUNNING: Once ovsdb-server is up and ovs-vswitchd has run the startup script. - UPDATE: Once a new pod is created and ovsdb-server-init has run. - RESTART_VSWITCH: After ovsdb-server-init has finished, new ovsdb-server pod has stopped the old ovs-vswitchd process. - RESTART_DBSERVER: After old ovs-vswitchd has been restarted the old ovsdb-server is also stop. The normal flow of states is the following: NULL -> INIT -> OVSDB_SERVER -> RUNNING Scale down: If the oc worker is deleted the DS and all the pods and mount points will be deleted, in case of node being up again it should start from NULL Update: RUNNING -> (Change on CR) -> UPDATE -> RESTART_VSWITCHD -> RESTART_DBSERVER -> OVSDB_SERVER -> RUNNING Jira: OSPRH-11636 Related: OSPRH-10821
|
Build failed (check pipeline). Post https://softwarefactory-project.io/zuul/t/rdoproject.org/buildset/fada0b90762848388e290079e5af88e4 ✔️ openstack-k8s-operators-content-provider SUCCESS in 1h 21m 43s |
|
@averdagu: The following test failed, say
Full PR test history. Your PR dashboard. 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-sigs/prow repository. I understand the commands that are listed here. |
This commit aims to modify the starting scripts of the ovn-controller-ovs daemonset.
This is done to allow modifying the RollingUpdate Strategy to not allow any Unavailable pod during update, what this will cause is that during an update to the pod (delete old one and create new one) instead of first deleting the first one and then deleting the next one, it will create the new pod while the old one is running. Due to how the startup scritps works currently this is not allowed.
The reason behind this is to try to lower the downtime observed if the environment is using centralized floating ip during an update.
With this commit the ovn-controller-ovs will share the PID namespace with the host, in order to allow signaling between old/new pod.
Another change is adding an STATE that the containers (ovsdb-server, ovs-vswitchd and ovsdb-server-init) will handle internally.
The differents states are:
The normal flow of states is the following:
NULL -> INIT -> OVSDB_SERVER -> RUNNING
Scale down: If the oc worker is deleted the DS and all the pods and mount points will be deleted, in case of node being up again it should start from NULL
Update: RUNNING -> (Change on CR) -> UPDATE -> RESTART_VSWITCHD ->
RESTART_DBSERVER -> OVSDB_SERVER -> RUNNING
Related: OSPRH-11636
Jira: OSPRH-10821
Depends-on: lib-common#611