perf(l1): enable unordered writes for trie node column families

[dicethedev]

Motivation
Trie node writes during block execution do not require strict ordering guarantees. RocksDB's default ordered-write path adds unnecessary synchronization overhead for these CFs.

Description
Adds cf_opts.set_unordered_write(true) to ACCOUNT_TRIE_NODES and STORAGE_TRIE_NODES. Unordered writes allow RocksDB to skip WAL sequencing constraints, potentially increasing write throughput for state-heavy workloads.

Not applied to other CFs where ordering or consistency guarantees may matter (e.g. HEADERS, BODIES, RECEIPTS).

Checklist

• Updated STORE_SCHEMA_VERSION (crates/storage/lib.rs) if the PR
  includes breaking changes to the Store requiring a re-sync.

Closes #5937

[greptile-apps]

Greptile Summary

This PR adds cf_opts.set_unordered_write(true) to the ACCOUNT_TRIE_NODES and STORAGE_TRIE_NODES column family configurations, aiming to skip WAL sequencing overhead for trie writes and improve write throughput.

• The change is a no-op: unordered_write is a DBOptions field in RocksDB, not a ColumnFamilyOptions field. When set on options passed to a ColumnFamilyDescriptor, RocksDB silently ignores it — the value is never read during CF initialization. The intended performance gain will not be realized.
• To actually enable this, it must be set on the top-level opts (line 32), which would apply it DB-wide (per-CF granularity is not supported). That path would also need reconciliation with enable_pipelined_write(true) and the PointInTime WAL recovery mode already configured.

Confidence Score: 3/5

Safe to merge in terms of stability (the option is ignored), but the claimed performance optimization is not achieved — the PR does not do what it intends.

There is one P1 finding: set_unordered_write(true) is silently a no-op when applied to CF-level options, so the PR's stated goal is not met. No crash or data-corruption risk from the current change, but shipping it as-is gives a false impression that the optimization is active.

crates/storage/backend/rocksdb.rs — specifically line 133 and the relationship between the CF-level cf_opts and the DB-level opts at line 32.

Important Files Changed

Filename	Overview
crates/storage/backend/rocksdb.rs	Adds set_unordered_write(true) to trie node CF options, but this is a no-op because unordered_write is a DB-level (DBOptions) setting in RocksDB that is silently ignored when applied to CF-level options in a ColumnFamilyDescriptor.

Flowchart

%%{init: {'theme': 'neutral'}}%%
flowchart TD
    A[RocksDB open] --> B[DB-level opts\nenable_pipelined_write=true\nwal_recovery_mode=PointInTime\nunordered_write=NOT SET]
    A --> C[CF loop]
    C --> D{CF name?}
    D --> |ACCOUNT_TRIE_NODES\nSTORAGE_TRIE_NODES| E[cf_opts\nset_unordered_write=true\n⚠️ DB-level option, silently ignored]
    D --> |HEADERS / BODIES / etc.| F[cf_opts\nstandard settings]
    E --> G[ColumnFamilyDescriptor]
    F --> G
    G --> H[open_cf_descriptors\nunordered_write flag dropped,\nnever applied to DB write path]
    B --> H
    H --> I[RocksDB running\nno unordered_write effect]

Loading

Prompt To Fix All With AI

This is a comment left during a code review.
Path: crates/storage/backend/rocksdb.rs
Line: 133

Comment:
**`unordered_write` is a DB-level option, not a CF-level option — this is a no-op**

`unordered_write` belongs to `DBOptions` in RocksDB, not `ColumnFamilyOptions`. When it is set on the `Options` object passed to a `ColumnFamilyDescriptor`, the flag is silently ignored during CF-open because RocksDB only reads the DB-level portion of the options from the primary DB-open call, not from individual CF descriptors (confirmed by the [RocksDB blog](http://rocksdb.org/blog/2019/08/15/unordered-write.html), which shows it set exclusively on `DBOptions`).

To actually take effect, `set_unordered_write(true)` must be called on the top-level `opts` object (line 32). However, doing so makes it a DB-wide setting — RocksDB has no mechanism to enable unordered writes selectively per column family. Additionally, enabling it at DB level would need verification against `enable_pipelined_write(true)` (line 60) and the `PointInTime` WAL recovery mode (line 54), which may require configuration adjustments.

How can I resolve this? If you propose a fix, please make it concise.

_{Reviews (1): Last reviewed commit: "perf(l1): enable unordered writes for tr..." | Re-trigger Greptile}

[greptile-apps]

unordered_write is a DB-level option, not a CF-level option — this is a no-op

unordered_write belongs to DBOptions in RocksDB, not ColumnFamilyOptions. When it is set on the Options object passed to a ColumnFamilyDescriptor, the flag is silently ignored during CF-open because RocksDB only reads the DB-level portion of the options from the primary DB-open call, not from individual CF descriptors (confirmed by the RocksDB blog, which shows it set exclusively on DBOptions).

To actually take effect, set_unordered_write(true) must be called on the top-level opts object (line 32). However, doing so makes it a DB-wide setting — RocksDB has no mechanism to enable unordered writes selectively per column family. Additionally, enabling it at DB level would need verification against enable_pipelined_write(true) (line 60) and the PointInTime WAL recovery mode (line 54), which may require configuration adjustments.

Prompt To Fix With AI

This is a comment left during a code review.
Path: crates/storage/backend/rocksdb.rs
Line: 133

Comment:
**`unordered_write` is a DB-level option, not a CF-level option — this is a no-op**

`unordered_write` belongs to `DBOptions` in RocksDB, not `ColumnFamilyOptions`. When it is set on the `Options` object passed to a `ColumnFamilyDescriptor`, the flag is silently ignored during CF-open because RocksDB only reads the DB-level portion of the options from the primary DB-open call, not from individual CF descriptors (confirmed by the [RocksDB blog](http://rocksdb.org/blog/2019/08/15/unordered-write.html), which shows it set exclusively on `DBOptions`).

To actually take effect, `set_unordered_write(true)` must be called on the top-level `opts` object (line 32). However, doing so makes it a DB-wide setting — RocksDB has no mechanism to enable unordered writes selectively per column family. Additionally, enabling it at DB level would need verification against `enable_pipelined_write(true)` (line 60) and the `PointInTime` WAL recovery mode (line 54), which may require configuration adjustments.

How can I resolve this? If you propose a fix, please make it concise.