Release 3.2.6 #3727
Merged
Release 3.2.6 #3727
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
karol-kokoszka
force-pushed
the
release-3.2.6
branch
from
22a7140 to
2481b41
Compare
|
@karol-kokoszka this branch is still missing some commits, e.g. f1aef7d - results in incorrect test env setup in gh actions: |
|
looks I have to cherry pick the commits again, it will be easier. Last non-docs related included is 55fbc26 |
Bumps [golang.org/x/crypto](https://github.com/golang/crypto) from 0.14.0 to 0.17.0. - [Commits](golang/crypto@v0.14.0...v0.17.0) --- updated-dependencies: - dependency-name: golang.org/x/crypto dependency-type: direct:production ... Signed-off-by: dependabot[bot] <[email protected]>
This test used to be failing with Scylla 5.4 and IPV6 because of different IPV6 string representations.
There are some differences in schema propagation in Scylla 5.4.0 with consistent_cluster_management enabled: scylladb/scylladb#16349 With this setup, given scenario could fail: - backup dc1 to loc1 and dc2 to loc2 - restore schema only from loc1 (so restore schema only to dc1) The result could be that nodes from dc1 have new, correct schema, while nodes from dc2 have the same schema as before the restore. For this reason, it's safer to download schema from every location to every node with location access. This does not solve the whole issue, but makes restore schema more likely to succeed.
Since Scylla 5.4 is out, and we support only 2 recent minor releases, we don't need to test SM against Scylla 5.1 anymore.
Since we support cluster with and without raft schema changes, we should test both of those cases (especially for restore schema tests). Fixes #3644
With this commit it's possible to run start-dev-env SKIP_GOSSIP=true which results in adding 'skip_wait_for_gossip_to_settle: 0' to scylla.yaml. Recently improved testing cluster setup seems to work fine (and faster) with that. However, not waiting for gossip breaks restore schema tests. This means that if someone wants to work on anything except restore schema, they could use this option to make testing cluster setup way faster.
Since StorageServiceRepairStatus (without timeout param) returns only when the repair job has finished, we shouldn't time out on our end (even if backoff retry could handle that).
This resulted in many backoff errors in SM logs even on successful repair:
{"L":"INFO","T":"2023-12-01T01:31:54.398Z","N":"cluster.client","M":"HTTP retry backoff","operation":"StorageServiceRepairStatus","wait":"28.607063257s","error":"after 16m0s: context deadline exceeded","_trace_id":"uqSjtDSfRoOl1WhetiLPgA"}
Since StorageServiceSstablesByKeyspacePost returns only when load&stream has finished, we shouldn't time out on our end (even if backoff retry could handle that).
This resulted in many backoff errors in SM logs even on successful restore:
{"L":"INFO","T":"2023-11-30T23:03:07.117Z","N":"cluster.client","M":"HTTP retry backoff","operation":"StorageServiceSstablesByKeyspacePost","wait":"999.175032ms","error":"after 30s: context deadline exceeded","_trace_id":"uqSjtDSfRoOl1WhetiLPgA"}
There are still some bugs (one described in the issue below) regarding using float intensity. In order to get rid of them, we should only tolerate float intensity at the entrypoint, but completely remove it from SM repair internals (both in service and progress display). We keep float intensity in task properties and in swagger endpoint parameters, but we convert it to int intensity as a first thing in internals, which happens in: - GetTarget (from task properties) - sctool repair control endpoint (from query param) Fixes #3665
Also, add timeout on first node setup, as misconfiguration could lead to hanging at this step.
Newer Scylla versions (e.g. 2024) docker images don't run ssh server on them own, but we require it for some of SM tests.
Because of problems with restoring backups into Scylla 5.4 with raft schema enabled (#3662), we want to test the following workaround: - use fresh cluster without raft schema - restore as usual - enable raft schema in the cluster In order to do that, we leave raft schema on src cluster and test how it works with raft schema enabled/disabled on dst cluster.
Removing filtering was done so that our tests can pass with Scylla 5.4 and raft enabled, but it didn't improve the real life situations where agents don't have cross region remote location access.
…ing CQL SSL Previously, SSL was preferred when client_encryption_options.enabled coming from ScyllaDB configuration was true and SSL port is open, even when Scylla Manager did not have any client certificate registered for particular cluster. This caused issues when ScyllaDB cluster was exposing both CQL and CQL SSL with mTLS, because even when Manager was not registered with certificates, it still insisted to establish sessions using SSL port. CQL healthchecks was also affected. Fixes #3698
"/storage_service/describe_ring/" returns the token range of a
random keyspace. this API is never used in production. what is used
is is cousin which accepts a mandatory keyspace path parameter, like
"/storage_service/describe_ring/{keyspace}", so let's drop it in
scylla-manager, we will drop this API in scylla as well.
Signed-off-by: Kefu Chai <[email protected]>
…use non SSL port on cluster Manager 3.2.6 gives a possibility of explicitly disabling TLS on session even though the certificate and key are available `force_tls_disabled`. Besides that, there is an option to force session to always use non-tls port from scylla config `force_non_ssl_session_port`.
…bled and --force-non-ssl-session-port flags
…on from DB This addresses #3679 .
It adds information about `force-tls-disabled` and `force-non-ssl-session-port` flags.
…-port cluster properties
The fix for #3707 includes extending the cluster kept in the DB with the 'host' property, which represents the initial host used when adding a cluster to Scylla Manager. Scylla-operator passes the DNS name here, making it immune to the ephemeral IPs in the Kubernetes environment. This commit adds the initial host as the first host to query in order to discover node IPs.
…ndpoint This endpoint returns 'consistent_cluster_management' option from scylla.yaml.
karol-kokoszka
force-pushed
the
release-3.2.6
branch
from
78dee7d to
63f3c8a
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Includes the following:
Please make sure that: