Added upgrade mechanism for managed Jaeger instances

[jpkrohling]

This PR is still in draft mode. The purpose for now is to agree on the main assumptions, like versioning and how to apply the upgrades.

Closes #42

Signed-off-by: Juraci Paixão Kröhling [email protected]

[codecov]

Codecov Report

Merging #476 into master will decrease coverage by 0.16%.
The diff coverage is 62.5%.

@@            Coverage Diff             @@
##           master     #476      +/-   ##
==========================================
- Coverage    91.7%   91.53%   -0.17%     
==========================================
  Files          64       65       +1     
  Lines        3181     3189       +8     
==========================================
+ Hits         2917     2919       +2     
- Misses        184      188       +4     
- Partials       80       82       +2

Impacted Files	Coverage Δ
pkg/apis/jaegertracing/v1/jaeger_types.go	100% <ø> (ø)	⬆️
pkg/controller/jaeger/jaeger_controller.go	38.14% <0%> (-2.07%)	⬇️
pkg/deployment/all-in-one.go	100% <100%> (ø)	⬆️
pkg/deployment/collector.go	100% <100%> (ø)	⬆️
pkg/deployment/ingester.go	100% <100%> (ø)	⬆️
pkg/deployment/agent.go	100% <100%> (ø)	⬆️
pkg/deployment/query.go	100% <100%> (ø)	⬆️
pkg/storage/cassandra_dependencies.go	100% <100%> (ø)	⬆️
pkg/controller/jaeger/upgrade.go	50% <50%> (ø)

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 35880d0...5cc91ab. Read the comment docs.

[codecov]

Codecov Report

Merging #476 into master will decrease coverage by 0.42%.
The diff coverage is 64.13%.

@@            Coverage Diff             @@
##           master     #476      +/-   ##
==========================================
- Coverage   91.98%   91.56%   -0.43%     
==========================================
  Files          69       73       +4     
  Lines        3557     3615      +58     
==========================================
+ Hits         3272     3310      +38     
- Misses        199      213      +14     
- Partials       86       92       +6

Impacted Files	Coverage Δ
pkg/deployment/args.go	100% <ø> (ø)	⬆️
pkg/apis/jaegertracing/v1/jaeger_types.go	100% <ø> (ø)	⬆️
pkg/upgrade/versions.go	100% <100%> (ø)
pkg/deployment/all-in-one.go	100% <100%> (ø)	⬆️
pkg/deployment/collector.go	100% <100%> (ø)	⬆️
pkg/deployment/ingester.go	100% <100%> (ø)	⬆️
pkg/deployment/agent.go	100% <100%> (ø)	⬆️
pkg/deployment/query.go	100% <100%> (ø)	⬆️
pkg/upgrade/v1_13_1.go	100% <100%> (ø)
pkg/storage/cassandra_dependencies.go	100% <100%> (ø)	⬆️
... and 8 more

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 584224f...f851e08. Read the comment docs.

[objectiser]

@jpkrohling Wanted to check one thing - currently it appears as if the jaeger version can be updated while the operator is running, as the controller checks for upgrading the version when a change is applied? Thought this would only be done when the operator is first started?

How would the version be updated for a running instance?

[jpkrohling]

The usual case will be that the operator will get all the existing Jaeger instances and upgrade them, but there might be some concurrency-related situations (or permission-related cases) where the operator couldn't see an existing Jaeger instance when it first started. The reconcile logic will catch these cases and upgrade the individual instances on demand.

How would the version be updated for a running instance?

Pretty much all Jaeger instances will be running when the upgrade process starts. Each upgrade can define how the upgrade will work. In some cases, we might need to stop the storage, apply a new schema or data migration, and start the collector again. In other cases, we can just change the deployment to use a newer image and Kubernetes will take care of the rolling update.

[jpkrohling]

@objectiser : have you had a chance to look into this PR? Do you think this would be appropriate for our planned use cases?

[objectiser]

@jpkrohling Yes I've added a number of comments that you have responded to :) Yes looks a long the right lines for what we need. Not sure should be added for 1.13.x yet though - possibly plan for 1.14?

[jpkrohling]

The coverage is lacking because it's not easy to simulate client problems with the fake client.