Feature branch for Config Cache Service

karol-kokoszka commented 5 months ago

Fixes https://github.com/scylladb/scylla-manager/issues/3767

PR is ready for a review.

It consists of the following commits:

https://github.com/scylladb/scylla-manager/pull/3816/commits/2b77a40688d8e975ed0ab8858bbe6897267c5e6c :This commit addresses the lack of an interface definition for the cluster service, which consists of CRUD operations on the cluster object.
https://github.com/scylladb/scylla-manager/pull/3816/commits/7947bc3893ec8fb6fe738fb09c2970894fee702f: Contains the implementation of the config cache service.
https://github.com/scylladb/scylla-manager/pull/3816/commits/303df330a45ad44d5061f81985c33cb101858b69: This commit defines the contract of the epic - to make the CQL ping and Alternator ping independent from the agent's API responsiveness.
https://github.com/scylladb/scylla-manager/pull/3816/commits/26517717830eee0c6a1a8ed0b06a05b0d73e32bd: Includes the config cache service into the application. It is initialized and started in the background.
https://github.com/scylladb/scylla-manager/pull/3816/commits/e1cfdf2fca5dbc4a7d0fd504ad137a27591c5807: Changes the Healthcheck service to start consuming data from the config cache service.
https://github.com/scylladb/scylla-manager/pull/3816/commits/f8b30d7be1a0f483b49d6e34951f1926aa537b5c: This commit removes the old caching approach from the Healthcheck service.
https://github.com/scylladb/scylla-manager/pull/3816/commits/61ba6397a0c405d127c2feae6a5b5e6b9bc652d7: Another cleaning commit that simplifies the logic used to create the TLS configuration for CQL and Alternator sessions.

Open discussion: Does it make sense to merge the PR by rebasing it to master, or should we rather squash and merge (I'm starting to lean towards the latter)?

How was it tested ?

Additional integration test which blocks the agent's HTTP API on one of the nodes and validates if CQL and Alternator pings on this node will succeed.

SCT sanity check executed against this branch:

CQL healtcheck available all the time:

test_client_encryption shows what happens when the node configuration changes

CQL is being unavailable for some time, until the configuration is refreshed and then it's reporting correct times again.

Please make sure that:

Code is split to commits that address a single change
Commit messages are informative
Commit titles have module prefix
Commit titles have issue nr. suffix

karol-kokoszka commented 5 months ago

Created DraftPR to continuously integrate the code to feature branch (I can observe integration tests).

karol-kokoszka commented 5 months ago

--- FAIL: TestServiceRepairIntegration/drop_keyspace_during_repair keeps failing (although no changes to anything other than healtcheck is introduced in this PR)

karol-kokoszka commented 4 months ago

@Michal-Leszczynski I'm merging the PR. Thanks for the review ! The SCT sanity passes on this branch.

scylladb / scylla-manager

Feature branch for Config Cache Service #3816