MWRS Status Expectations

[annelaucg]

TLDR: What is the expectation for the existing status.Conditions and status.Summary. See below for specific questions on each status.

This is the current state that I see:

• PlacementVerified
  • False: (PlacementDecisionNotFound, PlacementDecisionEmpty) --> This indicates that MWRS could not find a placement to refer to.
  • True: (AsExpected) --> This indicates that MWRS has found a placement to refer to successfully.
• PlacementRolledOut (True == all clusters rolled out)
  • False: (Progressing). This indicates that the MWRS is still rolling out to clusters.
  • True: (Completed). This indicates that the MWRS has been rolled out to all clusters.
  • Q: Should this only show True when successfully deployed?
• ManifestWorkApplied (True == ManifestWork is successfully applied to a cluster)
  • False: (Processing, NotAsExpected). This indicates that the ManifestWork is still being applied to a cluster or has status.Degraded == False, or some other issue with applying the ManifestWork.
  • True: (AsExpected). This indicates that the current ManifestWork is status.Progressing == False + status.Degraded == False /NotSet
  • Q: Should this only show True when all ManifestWorks are successfully deployed?

Based on the existing status, the only way to track if the entire MWRS is completed is to check if PlacementRolledOut == True and ManifestWorkApplied == True.

I also have questions about the following status.Summary values:

1. available ==> this refers to the current available ManifestWork.
   a. This number shows if the resources created by the ManifestWork is present on the cluster, not if the ManifestWork is applied successfully.
   b. Is the expectation that available can be aggregated before progressing is done?
2. total ==> this refers to the total number of ManifestWork currently deployed.
   a. Not the total clusters the placement selects, instead it shows the current deployed.
   b. Is this expected behavior?
   c. If it is, then in the check on how to set ManifestworkApplied == AsExpected, does this only show if the CURRENT Manifestwork is completed, but not all of the works? Here is the current code with the check that summary.total == summary.available and summary.progressing == 0 and summary.degraded == 0.

Ideally I want to track the following things:

1. How many clusters have been rolled out?
   a. Currently there is no way to get the total clusters that the placement targets from the MWRS status.summary, we can get the current number from total OR available.
   b. Get the total clusters intended to rollout by viewing the placementSummary.availableDecisionGroups message.
2. Is the ManifestWorkApplied successfully deployed?
   a. I don't think this function exists on a per MW basis.
3. Is the entire MWRS completed?
   b. Track if PlacementRolledOut == True and ManifestWorkApplied == True.
   c. Should we create a Ready/Progressing status to track this? Enhancement proposal here.

[dosubot]

Here’s how MWRS status conditions and summary fields work:

• PlacementVerified is True only when a valid placement is found; False means the placement is missing or empty (docs).
• PlacementRolledOut is True only when all clusters selected by the placement have ManifestWorks rolled out; otherwise, it’s False (Progressing). This does not guarantee successful deployment, just that rollout is complete (code).
• ManifestWorkApplied is True only when all ManifestWorks are successfully applied (i.e., summary.total == summary.available, summary.progressing == 0, summary.degraded == 0). If any ManifestWork is still progressing or degraded, it’s False (code).
• status.summary.available counts clusters where resources from ManifestWorks are present, not necessarily successfully applied. This can be aggregated before all are done progressing (docs).
• status.summary.total reflects the number of ManifestWorks currently deployed, not the total clusters targeted by placement. This is expected behavior (docs).
• To track overall MWRS completion, check both PlacementRolledOut == True and ManifestWorkApplied == True (docs).
• The total clusters intended for rollout can be found in placementSummary.availableDecisionGroups, not in MWRS status.summary (docs).
• There’s currently no per-ManifestWork “Applied” status in MWRS; it’s aggregated at the MWRS level (docs).
• Adding a Ready/Progressing status for more granular tracking is suggested in the enhancement proposal, but not yet implemented (discussion).

Let me know if you want details on edge cases or rollout strategy limitations.

To reply, just mention @dosu.

---

^{How did I do? Good | Irrelevant | Incorrect | Verbose | Hallucination | Report 🐛 | Other}    

[mikeshng]

CC @jnpacker @haoqing0110 @elgnay @zhujian7 @bhperry @zhiweiyin318

[annelaucg]

The total clusters intended for rollout can be found in placementSummary.availableDecisionGroups, not in MWRS status.summary (docs).

What is the reason for this behavior?

[qiujian16]

Q: Should this only show True when all ManifestWorks are successfully deployed?

When all manifestwork are applied. It depends. Controller will periodically check if resource needs to be applied, and if needed, it will apply again. The apply action may fail due to various reason.

b. Is the expectation that available can be aggregated before progressing is done?

it is supposed to be independent from progressing, it just intends to show the existence of the resource.

b. Is this expected behavior?

hrm, it might not be, I think we also want a desiredTotal which should be the total selected by the placement.

[qiujian16]

How many clusters have been rolled out?

+1

Is the ManifestWorkApplied successfully deployed?

I am not sure how to define successfully deployed.

Should we create a Ready/Progressing status to track this?

+1 for progressing, IMO ready is ambiguous, i think it will depend on workload to define what is ready?

[annelaucg]

When all manifestwork are applied. It depends. Controller will periodically check if resource needs to be applied, and if needed, it will apply again. The apply action may fail due to various reason.

The current behavior will set ManifestWorkApplied = True, when it has completed partial rollout. Eg. If ManifestWork applied on cluster1 and Progressing = False and Degraded = False or not set. So there is a bug here for the ManifestWorkApplied status.

hrm, it might not be, I think we also want a desiredTotal which should be the total selected by the placement.

So we want to keep the total field and create a new desiredTotal field to show the total? Is there a reason why total was changed to show the current selected by placement?

I am not sure how to define successfully deployed.
For this one I'm thinking about ManifestWork has Progressing=false and Degraded=False or not set.

Ready field that I am thinking is for letting end user note that the MWRS is 1. all on clusters 2. successfully running on cluster (based on conditionRules set)

[qiujian16]

So we want to keep the total field and create a new desiredTotal field to show the total? Is there a reason why total was changed to show the current selected by placement?

Hrm I do not think we change it before. Having both a number of "selected by placement" and "total manifestwork" might be valuable, since we can understand the ratio of deployed workload.

[annelaucg]

Update from our conversation in community call (Nov 6):

1. PlacementRolledOut should be updated to "True" only after ManifestWorkApplied is completed for all clusters.
2. Create enhancement for Progressing status and add a desiredTotal field in status.Summary that shows the true total selected by the placementRef.