d-wang-commit opened a new pull request, #18692:
URL: https://github.com/apache/iotdb/pull/18692
## Description
### Policy use cases
The configuration template describes the intended scenarios as follows
(quoted verbatim).
SchemaRegionGroups:
```properties
# 1. CUSTOM: For known workloads that need manual resource allocation;
creates the configured target
# number of SchemaRegionGroups per Database when schema partitions are
first allocated.
# 2. AUTO: For most routine workloads (roughly 80% as an approximate
planning guideline); gradually
# meets the configured minimum and expands with slot occupancy up to the
resource-based maximum,
# balancing parallelism and RegionGroup management overhead.
# 3. PROACTIVE: Complements AUTO where it does not expand eagerly enough,
such as very few devices with
# many measurements and a high load; creates groups earlier for more
parallelism at the cost of
# additional RegionGroup management overhead.
```
DataRegionGroups:
```properties
# 1. CUSTOM: For known workloads that need manual resource allocation;
creates the configured target
# number of DataRegionGroups per Database when data partitions are first
allocated.
# 2. AUTO: For most routine workloads (roughly 80% as an approximate
planning guideline); gradually
# meets the configured minimum and expands with slot occupancy up to the
resource-based maximum,
# balancing parallelism and RegionGroup management overhead.
# 3. PROACTIVE: Complements AUTO where it does not expand eagerly enough,
such as very few devices with
# many measurements and a high load; creates groups earlier for more
parallelism at the cost of
# additional RegionGroup management overhead.
```
### Proactive RegionGroup allocation
Sparse databases can keep using only a few RegionGroups under AUTO even when
resources are available. Add PROACTIVE for SchemaRegionGroups and
DataRegionGroups, and make it the default in ConfigNodeConfig and the
configuration template.
PROACTIVE counts distinct persisted and pending series slots separately for
schema and data. It grows toward one group per active slot, respects the
resource-derived maximum, and follows AUTO's incremental growth toward the
configured minimum. New time partitions do not inflate the distinct-slot count.
Existing groups are retained; if all groups of the requested type are disabled,
one replacement can be allocated within the maximum.
For data allocation and recovery, calculate the retained-slot limit from
active assignments, with a minimum of one, instead of the total configured slot
count. With 1,000 slots and six active devices, switching from AUTO with two
groups to PROACTIVE can grow to six groups: each old group retains one slot and
four slots use the four new groups for new partitions. Historical partition
mappings remain unchanged.
### Policy hot reload
Recalculate resource-derived maxima on the ConfigNode leader when either
extension policy changes, as well as when per-node quotas change. Previously,
leaving CUSTOM without changing quotas could retain its manual cap and prevent
expected growth.
Cover CUSTOM below and above the PROACTIVE target, PROACTIVE to CUSTOM,
CUSTOM to AUTO, and independent schema/data switching. Switching to CUSTOM
preserves existing groups and uses the stored maximum; it does not promise to
redistribute sparse slots across every group.
### Tests
Added UTs:
- PartitionManagerProactiveExtensionTest: distinct slots, minimum growth,
batching, caps, unavailable groups, policy switches, and failed-allocation
retries.
- DataPartitionPolicyTableProactiveTest: sparse assignment, use of new
groups, recovery, and AUTO/CUSTOM compatibility.
- RegionGroupExtensionPolicyConfigTest: parsing, invalid values, file
loading, and independent hot reload.
Existing ClusterSchemaManagerTest and DataPartitionPolicyTableTest provide
resource-limit and allocation regression coverage.
Added ITs:
- IoTDBProactiveRegionGroupDistributionIT: AUTO/PROACTIVE comparisons on
1C3D, 1C6D, and 1C9D; replication, SQL reads/writes, hot reload, new time
partitions, and ConfigNode restart. Six targeted hot-reload cases cover
AUTO/CUSTOM/PROACTIVE transitions and preserved historical mappings.
- IoTDBProactiveRegionGroupExtensionIT: five topology/consensus combinations
covering partition counts, shared slots, caps, ConfigNode recovery, and mapping
retention.
The recorded local targeted run completed with 65 UT cases (0 failures, 0
errors, 3 skipped) and all six hot-reload IT cases passing. The topology matrix
and existing IoTDBPartitionInheritStrategyIT were not part of that run.
### Side effects and risks
Changing the default from AUTO to PROACTIVE can create more RegionGroups
earlier, increasing consensus, memory, and storage overhead within resource
limits. Explicit AUTO and CUSTOM settings remain available. Policy changes can
affect placement of newly allocated partitions, but do not migrate historical
data or shrink existing groups. Allocation may still fail when insufficient
DataNodes are available.
<!--
In each section, please describe design decisions made, including:
- Choice of algorithms
- Behavioral aspects. What configuration values are acceptable? How are
corner cases and error
conditions handled, such as when there are insufficient resources?
- Class organization and design (how the logic is split between classes,
inheritance, composition,
design patterns)
- Method organization and design (how the logic is split between methods,
parameters and return types)
- Naming (class, method, API, configuration, HTTP endpoint, names of
emitted metrics)
-->
<!-- It's good to describe an alternative design (or mention an alternative
name) for every design
(or naming) decision point and compare the alternatives with the designs
that you've implemented
(or the names you've chosen) to highlight the advantages of the chosen
designs and names. -->
<!-- If there was a discussion of the design of the feature implemented in
this PR elsewhere
(e. g. a "Proposal" issue, any other issue, or a thread in the development
mailing list),
link to that discussion from this PR description and explain what have
changed in your final design
compared to your original proposal or the consensus version in the end of
the discussion.
If something hasn't changed since the original discussion, you can omit a
detailed discussion of
those aspects of the design here, perhaps apart from brief mentioning for
the sake of readability
of this PR description. -->
<!-- Some of the aspects mentioned above may be omitted for simple and small
changes. -->
<hr>
This PR has:
- [x] added documentation for new or modified features or behaviors.
- [x] added comments explaining the "why" and the intent of the code
wherever would not be obvious
for an unfamiliar reader.
- [x] added integration tests.
- [x] been tested in a test IoTDB cluster.
<!-- Check the items by putting "x" in the brackets for the done things. Not
all of these items
apply to every PR. Remove the items which are not done or not relevant to
the PR. None of the items
from the checklist above are strictly necessary, but it would be very
helpful if you at least
self-review the PR. -->
<hr>
##### Key changed/added classes (or packages if there are too many classes)
in this PR
- RegionGroupExtensionPolicy and ConfigNodeConfig: PROACTIVE policy and
defaults.
- PartitionManager, PartitionInfo, and DatabasePartitionTable: distinct-slot
counting and bounded proactive allocation.
- DataPartitionPolicyTable: active-slot-based retention during balancing and
recovery.
- ConfigManager: refresh resource-derived maxima after policy hot reload.
- New proactive/configuration UTs and proactive distribution/extension ITs:
allocation and switching coverage.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]