Update maxUnavailable calculation for leader StatefulSet

[adinilfeld]

What type of PR is this?

/kind feature

What this PR does / why we need it

This PR sets the maxUnavailable value of the leader StatefulSet to lws.maxSurge + lws.maxUnavailable, rather than just lws.maxUnavailable. This has two main benefits:

1. Users can now achieve zero-downtime updates by specifying maxUnavailable=0, as long as maxSurge>0. Previously, the leader StatefulSet would have rejected this value. (See Support maxUnavailable=0 #776.)

2. Rolling updates now update pods in larger batches, leading to rollouts completing quicker while still respecting the spec.

   For instance, the example given in the Rollout Strategy doc previously didn't work as documented. In Stage3, R-2 and R-3 would only begin updating after both surge pods were Ready, as the StatefulSet's maxUnavailable=2 was "used up" by the surge pods. Now it would have maxUnavailable=4 and be able to update as intended. In my testing, this change reduced the rollout time by 30% (from 1m48s to 1m15s).

In addition, I made some small tweaks to the logs for clarity.

Which issue(s) this PR fixes

Fixes #776.

Special notes for your reviewer

In addition to unit/e2e/integration tests, I verified the following scenarios manually:

maxUnavailable=0, maxSurge=2: once the two surge pods are Ready, the existing pods are updated two at a time.

Full events

3s          Normal   CreatingRevision                 leaderworkerset/lws-maxunavailable-0   Creating revision with key 6fdd78dd8d for updated LWS
3s          Normal   GroupsUpdating                   leaderworkerset/lws-maxunavailable-0   Rolling Upgrade is in progress, with 4 groups ready of total 4 groups
2s          Normal   GroupsUpdating                   leaderworkerset/lws-maxunavailable-0   Rolling Upgrade is in progress, with 4 groups ready of total 4 groups
3s          Normal   SuccessfulCreate                 statefulset/lws-maxunavailable-0       Create Pod lws-maxunavailable-0-5 in StatefulSet lws-maxunavailable-0 successful
3s          Normal   SuccessfulCreate                 statefulset/lws-maxunavailable-0       Create Pod lws-maxunavailable-0-4 in StatefulSet lws-maxunavailable-0 successful
2s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   Created worker statefulset for leader pod lws-maxunavailable-0-4
2s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   Created worker statefulset for leader pod lws-maxunavailable-0-5
0s          Normal   GroupsUpdating                   leaderworkerset/lws-maxunavailable-0   Updating replicas 2 to 3 (inclusive)
0s          Normal   SuccessfulDelete                 statefulset/lws-maxunavailable-0       Delete Pod lws-maxunavailable-0-3 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   SuccessfulDelete                 statefulset/lws-maxunavailable-0       Delete Pod lws-maxunavailable-0-2 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   SuccessfulCreate                 statefulset/lws-maxunavailable-0       Create Pod lws-maxunavailable-0-3 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   Created worker statefulset for leader pod lws-maxunavailable-0-3
0s          Normal   SuccessfulCreate                 statefulset/lws-maxunavailable-0       Create Pod lws-maxunavailable-0-2 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   Created worker statefulset for leader pod lws-maxunavailable-0-2
0s          Normal   GroupsUpdating                   leaderworkerset/lws-maxunavailable-0   Updating replica 1
0s          Normal   SuccessfulDelete                 statefulset/lws-maxunavailable-0       Delete Pod lws-maxunavailable-0-1 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   GroupsUpdating                   leaderworkerset/lws-maxunavailable-0   Updating replica 0
0s          Normal   SuccessfulDelete                 statefulset/lws-maxunavailable-0       Delete Pod lws-maxunavailable-0-0 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   SuccessfulCreate                 statefulset/lws-maxunavailable-0       Create Pod lws-maxunavailable-0-1 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   Created worker statefulset for leader pod lws-maxunavailable-0-1
0s          Normal   SuccessfulCreate                 statefulset/lws-maxunavailable-0       Create Pod lws-maxunavailable-0-0 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   Created worker statefulset for leader pod lws-maxunavailable-0-0
0s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   deleting surge replica lws-maxunavailable-0-5
0s          Normal   SuccessfulDelete                 statefulset/lws-maxunavailable-0       Delete Pod lws-maxunavailable-0-5 in StatefulSet lws-maxunavailable-0 successful
0s          Normal   GroupsProgressing                leaderworkerset/lws-maxunavailable-0   deleting surge replica lws-maxunavailable-0-4
0s          Normal   AllGroupsReady                   leaderworkerset/lws-maxunavailable-0   All replicas are ready, with 6 groups ready of total 4 groups
0s          Normal   SuccessfulDelete                 statefulset/lws-maxunavailable-0       Delete Pod lws-maxunavailable-0-4 in StatefulSet lws-maxunavailable-0 successful

maxUnavailable=2, maxSurge=2: the rollout order now matches the example given in the docs.

Full events

2m57s       Normal   CreatingRevision                 leaderworkerset/lws-rollout-doc-example   Creating revision with key 6fdd78dd8d for updated LWS
2m57s       Normal   GroupsUpdating                   leaderworkerset/lws-rollout-doc-example   Rolling Upgrade is in progress, with 4 groups ready of total 4 groups
2m57s       Normal   GroupsUpdating                   leaderworkerset/lws-rollout-doc-example   Updating replicas 2 to 3 (inclusive)
2m58s       Normal   SuccessfulCreate                 statefulset/lws-rollout-doc-example       Create Pod lws-rollout-doc-example-4 in StatefulSet lws-rollout-doc-example successful
2m58s       Normal   SuccessfulCreate                 statefulset/lws-rollout-doc-example       Create Pod lws-rollout-doc-example-5 in StatefulSet lws-rollout-doc-example successful
2m58s       Normal   SuccessfulDelete                 statefulset/lws-rollout-doc-example       Delete Pod lws-rollout-doc-example-3 in StatefulSet lws-rollout-doc-example successful
2m57s       Normal   GroupsProgressing                leaderworkerset/lws-rollout-doc-example   Created worker statefulset for leader pod lws-rollout-doc-example-4
2m58s       Normal   SuccessfulDelete                 statefulset/lws-rollout-doc-example       Delete Pod lws-rollout-doc-example-2 in StatefulSet lws-rollout-doc-example successful
2m57s       Normal   GroupsUpdating                   leaderworkerset/lws-rollout-doc-example   Rolling Upgrade is in progress, with 4 groups ready of total 4 groups
2m57s       Normal   GroupsProgressing                leaderworkerset/lws-rollout-doc-example   Created worker statefulset for leader pod lws-rollout-doc-example-5
2m54s       Normal   GroupsUpdating                   leaderworkerset/lws-rollout-doc-example   Updating replica 1
2m55s       Normal   SuccessfulDelete                 statefulset/lws-rollout-doc-example       Delete Pod lws-rollout-doc-example-1 in StatefulSet lws-rollout-doc-example successful
2m55s       Normal   SuccessfulDelete                 statefulset/lws-rollout-doc-example       Delete Pod lws-rollout-doc-example-0 in StatefulSet lws-rollout-doc-example successful
2m26s       Normal   GroupsProgressing                leaderworkerset/lws-rollout-doc-example   Created worker statefulset for leader pod lws-rollout-doc-example-2
2m25s       Normal   GroupsProgressing                leaderworkerset/lws-rollout-doc-example   Created worker statefulset for leader pod lws-rollout-doc-example-3
2m23s       Normal   GroupsProgressing                leaderworkerset/lws-rollout-doc-example   Created worker statefulset for leader pod lws-rollout-doc-example-0
2m22s       Normal   GroupsProgressing                leaderworkerset/lws-rollout-doc-example   Created worker statefulset for leader pod lws-rollout-doc-example-1
104s        Normal   GroupsProgressing                leaderworkerset/lws-rollout-doc-example   deleting surge replicas from lws-rollout-doc-example-4 to lws-rollout-doc-example-5
105s        Normal   SuccessfulDelete                 statefulset/lws-rollout-doc-example       Delete Pod lws-rollout-doc-example-5 in StatefulSet lws-rollout-doc-example successful
105s        Normal   SuccessfulDelete                 statefulset/lws-rollout-doc-example       Delete Pod lws-rollout-doc-example-4 in StatefulSet lws-rollout-doc-example successful
101s        Normal   AllGroupsReady                   leaderworkerset/lws-rollout-doc-example   All replicas are ready, with 6 groups ready of total 4 groups

Does this PR introduce a user-facing change?

NONE

[netlify]

✅ Deploy Preview for kubernetes-sigs-lws canceled.

Name	Link
🔨 Latest commit	0aa7d94
🔍 Latest deploy log	https://app.netlify.com/projects/kubernetes-sigs-lws/deploys/69c315aa19fd730008fc2d93

[adinilfeld]

/test pull-lws-test-integration-main

[adinilfeld]

/test pull-lws-test-integration-main

[Edwinhr716]

/lgtm

Will approve once #778 is merged

[adinilfeld]

/test pull-lws-test-integration-main

[Edwinhr716]

/lgtm
/approve

[k8s-ci-robot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: adinilfeld, Edwinhr716

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Details

Needs approval from an approver in each of these files:

• OWNERS [Edwinhr716]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment