storage: enable space management by default #15996
Conversation
There was a problem hiding this comment.
Space management is now enabled by default for upgraded clusters starting in v23.3, unless the feature has been explicitly disabled.
I'm a bit confused by the mention of v23.3 here. Does this mean "clusters that were bootstrapped on versions higher than v23.3 will enable space management automatically" or "all clusters that exist in v23.3 will enable space management automatically, regardless of upgrades"? If the latter, perhaps remove the mention of v23.3; the release note would be applicable to any version this is backported to.
Overall looks non-controversial, just wanna clarify the versioning expectations. Perhaps an upgrade test would help clarify the intent?
|
ducktape was retried in https://buildkite.com/redpanda/redpanda/builds/43574#018ceb11-6778-43e2-94f6-0267ecf8d8e0 |
dotnwat
force-pushed
the
default-sm-enable
branch
from
7cf2ae8
to
2d5f810
Compare
The following configuration options have these default values regardless of the upgrade history of the cluster space_management_enable=true retention_local_strict=false Additionally these options used to work around an issue in v23.2 which prevented changes to the above options from being sticky in an upgraded cluster no longer have any affect. space_management_enable_override retention_local_strict_override They'll be fully deprecated at a later time. Signed-off-by: Noah Watkins <noahwatkins@gmail.com>
| # in 23.2 space management exists, but is disabled by default for | ||
| # upgraded clusters. | ||
| self._upgrade(wipe_cache, self.intermediate_version) | ||
| self._check_value_everywhere("space_management_enable", False) | ||
| self._check_value_everywhere("retention_local_strict", True) | ||
|
|
||
| # in >=23.3 (using upstream build) space management should be enabled by | ||
| # default provided that it wasn't explicitly disabled in 23.2. in this | ||
| # case no configs were changed so it should be enabled now. | ||
| self._upgrade(wipe_cache) | ||
| self._check_value_everywhere("space_management_enable", True) | ||
| self._check_value_everywhere("retention_local_strict", False) | ||
|
|
||
| # survives a restart | ||
| self.redpanda.restart_nodes(self.redpanda.nodes) | ||
| self._check_value_everywhere("space_management_enable", True) | ||
| self._check_value_everywhere("retention_local_strict", False) |
| # version where space_management legacy defaults are introduced | ||
| self.intermediate_version = (23, 2) |
There was a problem hiding this comment.
Unrelated to this PR, for some reason I was under the impression Redpanda would complain/fail if we skipped any logical versions. I guess not though, given these other tests are already jumping from 23.1 to 23.3-soon-to-be-24.1
There was a problem hiding this comment.
yeh good call out i think we should probably revisit the upgrade test infra if anything just to say that it still looks good
|
new failures in https://buildkite.com/redpanda/redpanda/builds/43620#018cf102-0352-4bee-80f5-6a33416de007: new failures in https://buildkite.com/redpanda/redpanda/builds/43620#018cf102-0348-43fd-80af-1c73b55a38be: new failures in https://buildkite.com/redpanda/redpanda/builds/43620#018cf102-034b-41f3-9921-109315fbafaf: new failures in https://buildkite.com/redpanda/redpanda/builds/43620#018cf112-1371-4dd5-91dd-7c47a5603614: new failures in https://buildkite.com/redpanda/redpanda/builds/43620#018cf112-136e-40fd-a13e-1621ecf8971a: new failures in https://buildkite.com/redpanda/redpanda/builds/43620#018cf112-136a-4d1d-8fe7-54e5c5d6859a: |
|
/backport v23.3.x |
Make the following configuration options have these default values regardless of the upgrade history of the cluster
space_management_enable=true
retention_local_strict=false
Additionally these options used to work around an issue in v23.2 which prevented changes to the above options from being sticky in an upgraded cluster no longer have any affect.
space_management_enable_override
retention_local_strict_override
They'll be fully deprecated at a later time.
Fixes: #15995
Backports Required
Release Notes
Improvements
Space management is now enabled by default for upgraded clusters starting in v23.3, unless the feature has been explicitly disabled.