Add HTTP remote partial failure tests

[timsehn]

Summary

• add an HTTP remotesrv test for failures after chunk upload but before refs publish
• assert retries work and the remote remains readable across failed refs-if and commit phases
• run the new test in non-Windows CI with the other HTTP remotesrv tests

Test

• make -C build -j$(sysctl -n hw.ncpu 2>/dev/null || nproc) DOLTLITE_PROLLY_CHECK=1 doltlite doltlite-remotesrv
• bash test/remotesrv_http_partial_failure_test.sh build/doltlite build/doltlite-remotesrv
• bash test/remotesrv_http_test.sh build/doltlite build/doltlite-remotesrv
• bash test/remotesrv_http_concurrency_test.sh build/doltlite build/doltlite-remotesrv

[github-actions]

Version-Control Performance Ceilings

Runs: median of 3 executions per benchmark, excluding fixture setup. The
working-set fixtures use 800 committed tables with 125 rows each
(100000 rows total); data-dirty cases update 1 row in 40
tables, and schema-dirty cases update 1 row plus add a column in 20 tables.
Branch tests use 300 branches, checkout uses a clean 400-table
branch switch over 100000 rows, and merge
tests use 100000-row tables with 2000 changed or conflicting
rows per side.

Benchmark	Median ms	Ceiling ms	Result
status_clean_many_tables	83	200	PASS
status_dirty_many_tables	84	200	PASS
diff_regular_working_one_table	77	150	PASS
diff_regular_working_many_tables	83	200	PASS
diff_stat_working_many_tables	85	200	PASS
diff_schema_working_many_tables	85	200	PASS
branch_list_many_branches	24	100	PASS
branch_create_delete	25	100	PASS
checkout_branch_clean	55	200	PASS
merge_data_no_conflicts	29	150	PASS
merge_schema_no_conflicts	21	100	PASS
merge_data_conflicts	127	250	PASS
merge_data_conflicts_with_resolve	126	250	PASS

[github-actions]

Sysbench-Style Benchmark: Doltlite vs SQLite

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	22,253	28,619	1.29
oltp_range_select	9,178	13,438	1.46
oltp_sum_range	8,816	11,581	1.31
oltp_order_range	2,435	2,830	1.16
oltp_distinct_range	3,494	3,947	1.13
oltp_index_scan	3,732	5,011	1.34
select_random_points	9,255	10,669	1.15
select_random_ranges	2,824	3,852	1.36
covering_index_scan	4,155	4,096	0.99
groupby_scan	29,279	31,786	1.09
index_join	5,925	7,929	1.34
index_join_scan	3,290	4,411	1.34
types_table_scan	1,038,442	1,267,393	1.22
table_scan	1,161,993	1,383,572	1.19
oltp_read_only	97,977	119,064	1.22
Average			1.24

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	174,981	245,922	1.41
oltp_insert	15,155	27,574	1.82
oltp_update_index	48,757	87,578	1.80
oltp_update_non_index	32,635	57,326	1.76
oltp_delete_insert	43,320	69,447	1.60
oltp_write_only	21,020	42,922	2.04
types_delete_insert	23,114	39,160	1.69
oltp_read_write	63,743	108,681	1.70
Average			1.73

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	91,021	52,011	0.57
oltp_range_select	16,697	16,211	0.97
oltp_sum_range	15,791	14,223	0.90
oltp_order_range	3,296	3,215	0.98
oltp_distinct_range	4,268	4,213	0.99
oltp_index_scan	10,653	7,668	0.72
select_random_points	16,568	13,543	0.82
select_random_ranges	9,661	6,364	0.66
covering_index_scan	11,332	6,537	0.58
groupby_scan	30,024	32,276	1.08
index_join	9,828	9,598	0.98
index_join_scan	4,320	4,879	1.13
types_table_scan	1,036,895	1,278,596	1.23
table_scan	1,164,415	1,390,893	1.19
oltp_read_only	197,492	154,966	0.78
Average			0.90

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	189,392	263,337	1.39
oltp_insert	21,702	34,760	1.60
oltp_update_index	74,232	112,776	1.52
oltp_update_non_index	55,688	80,045	1.44
oltp_delete_insert	65,480	91,251	1.39
oltp_write_only	43,278	61,377	1.42
types_delete_insert	38,523	52,182	1.35
oltp_read_write	86,733	127,712	1.47
Average			1.45

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	45,239	51,987	1.15
oltp_range_select	12,348	16,284	1.32
oltp_sum_range	11,449	14,305	1.25
oltp_order_range	2,786	3,215	1.15
oltp_distinct_range	3,865	4,347	1.12
oltp_index_scan	6,210	7,652	1.23
select_random_points	12,303	13,285	1.08
select_random_ranges	5,214	6,286	1.21
covering_index_scan	6,495	6,624	1.02
groupby_scan	29,424	32,575	1.11
index_join	7,278	9,668	1.33
index_join_scan	3,717	4,898	1.32
types_table_scan	1,032,864	1,276,413	1.24
table_scan	1,161,265	1,390,360	1.20
oltp_read_only	131,875	153,374	1.16
Average			1.19

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	20,953	81,790	3.90
oltp_insert_ac	24,412	105,185	4.31
oltp_update_index_ac	26,364	113,641	4.31
oltp_update_non_index_ac	21,892	90,939	4.15
oltp_delete_insert_ac	23,306	110,261	4.73
oltp_write_only_ac	23,798	99,620	4.19
types_delete_insert_ac	21,639	98,068	4.53
oltp_read_write_ac	28,896	111,856	3.87
Average			4.25

100000 rows, median of 5 invocations per test; autocommit writes use 9, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (2.5x individual, 2x average; autocommit writes: 6x / 5x)

All tests within ceilings.

[github-actions]

Sysbench-Style Benchmark (composite PK): Doltlite vs SQLite

Companion to the classic Sysbench-Style Benchmark. Every workload here
runs against tables with a 2-column INTEGER PRIMARY KEY(a, b) WITHOUT ROWID.
Individual ratios gated at 2.5×; section averages gated at 2×. Autocommit writes use 6× / 5×.

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	33,256	37,750	1.14
oltp_range_select	19,415	21,201	1.09
oltp_sum_range	17,280	19,786	1.15
oltp_order_range	3,708	3,888	1.05
oltp_distinct_range	4,709	4,973	1.06
oltp_index_scan	4,610	5,719	1.24
select_random_points	27,367	30,276	1.11
select_random_ranges	7,431	8,313	1.12
covering_index_scan	4,328	4,119	0.95
groupby_scan	37,989	40,923	1.08
index_join	7,983	10,078	1.26
index_join_scan	4,145	5,619	1.36
types_table_scan	1,140,829	1,261,125	1.11
table_scan	1,353,996	1,382,353	1.02
oltp_read_only	147,368	164,055	1.11
Average			1.12

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	245,253	339,828	1.39
oltp_insert	19,166	36,401	1.90
oltp_update_index	69,149	121,219	1.75
oltp_update_non_index	52,272	84,619	1.62
oltp_delete_insert	51,340	98,788	1.92
oltp_write_only	27,728	59,987	2.16
types_delete_insert	33,253	54,717	1.65
oltp_read_write	103,732	155,558	1.50
Average			1.74

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	119,474	67,997	0.57
oltp_range_select	29,627	24,063	0.81
oltp_sum_range	26,973	23,432	0.87
oltp_order_range	4,669	4,262	0.91
oltp_distinct_range	5,717	5,314	0.93
oltp_index_scan	13,441	9,235	0.69
select_random_points	38,597	34,737	0.90
select_random_ranges	16,217	11,573	0.71
covering_index_scan	13,159	7,741	0.59
groupby_scan	39,669	41,771	1.05
index_join	12,841	12,907	1.01
index_join_scan	5,240	6,470	1.23
types_table_scan	1,287,838	1,287,724	1.00
table_scan	1,546,407	1,437,803	0.93
oltp_read_only	276,762	211,046	0.76
Average			0.86

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	259,684	369,244	1.42
oltp_insert	26,546	47,866	1.80
oltp_update_index	101,713	155,879	1.53
oltp_update_non_index	80,221	109,107	1.36
oltp_delete_insert	79,754	124,806	1.56
oltp_write_only	51,117	82,973	1.62
types_delete_insert	50,777	70,043	1.38
oltp_read_write	128,344	176,373	1.37
Average			1.51

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	60,970	68,286	1.12
oltp_range_select	24,215	24,734	1.02
oltp_sum_range	21,451	23,672	1.10
oltp_order_range	4,177	4,306	1.03
oltp_distinct_range	5,157	5,361	1.04
oltp_index_scan	7,876	9,304	1.18
select_random_points	31,765	34,896	1.10
select_random_ranges	10,535	11,513	1.09
covering_index_scan	7,773	7,669	0.99
groupby_scan	39,207	41,580	1.06
index_join	10,194	12,882	1.26
index_join_scan	4,744	6,253	1.32
types_table_scan	1,186,868	1,253,569	1.06
table_scan	1,354,569	1,426,303	1.05
oltp_read_only	196,896	209,852	1.07
Average			1.10

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	18,456	63,155	3.42
oltp_insert_ac	17,906	84,840	4.74
oltp_update_index_ac	21,165	96,108	4.54
oltp_update_non_index_ac	18,939	79,945	4.22
oltp_delete_insert_ac	19,655	92,333	4.70
oltp_write_only_ac	18,703	87,662	4.69
types_delete_insert_ac	15,915	74,598	4.69
oltp_read_write_ac	25,244	97,924	3.88
Average			4.36

100000 rows, median of 5 invocations per test; autocommit writes use 9, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (2.5x individual, 2x average; autocommit writes: 6x / 5x)

All tests within ceilings.

[github-actions]

Sysbench-Style Benchmark (BLOB PK): Doltlite vs SQLite

Companion to the classic Sysbench-Style Benchmark. Every workload here
runs against tables with a 16-byte big-endian BLOB PRIMARY KEY.
Individual ratios gated at 2.5×; section averages gated at 2×. Autocommit writes use 6× / 5×.

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	32,493	37,656	1.16
oltp_range_select	13,014	14,739	1.13
oltp_sum_range	12,759	13,951	1.09
oltp_order_range	2,990	3,161	1.06
oltp_distinct_range	4,023	4,176	1.04
oltp_index_scan	4,701	6,417	1.37
select_random_points	18,504	20,452	1.11
select_random_ranges	4,204	5,279	1.26
covering_index_scan	4,389	4,619	1.05
groupby_scan	31,712	33,252	1.05
index_join	7,038	9,601	1.36
index_join_scan	4,134	5,522	1.34
types_table_scan	1,080,053	1,283,387	1.19
table_scan	1,347,700	1,436,966	1.07
oltp_read_only	127,630	138,829	1.09
Average			1.16

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	243,545	354,686	1.46
oltp_insert	20,156	39,045	1.94
oltp_update_index	73,368	133,848	1.82
oltp_update_non_index	51,290	86,220	1.68
oltp_delete_insert	52,128	104,665	2.01
oltp_write_only	29,191	61,262	2.10
types_delete_insert	34,655	54,570	1.57
oltp_read_write	93,368	141,305	1.51
Average			1.76

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	101,861	61,505	0.60
oltp_range_select	21,984	17,628	0.80
oltp_sum_range	19,800	16,552	0.84
oltp_order_range	4,032	3,688	0.91
oltp_distinct_range	5,021	4,668	0.93
oltp_index_scan	11,741	9,083	0.77
select_random_points	27,743	23,895	0.86
select_random_ranges	11,217	7,657	0.68
covering_index_scan	12,257	7,240	0.59
groupby_scan	33,538	34,109	1.02
index_join	11,141	11,376	1.02
index_join_scan	5,685	5,867	1.03
types_table_scan	1,110,571	1,301,464	1.17
table_scan	1,310,846	1,418,332	1.08
oltp_read_only	230,419	173,855	0.75
Average			0.87

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	260,294	375,402	1.44
oltp_insert	31,863	51,472	1.62
oltp_update_index	105,928	160,147	1.51
oltp_update_non_index	80,317	106,585	1.33
oltp_delete_insert	79,826	128,286	1.61
oltp_write_only	55,562	82,235	1.48
types_delete_insert	52,380	71,679	1.37
oltp_read_write	116,550	165,782	1.42
Average			1.47

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	55,924	60,947	1.09
oltp_range_select	17,121	17,490	1.02
oltp_sum_range	15,798	16,584	1.05
oltp_order_range	3,342	3,503	1.05
oltp_distinct_range	4,435	4,540	1.02
oltp_index_scan	7,283	8,909	1.22
select_random_points	23,059	24,095	1.04
select_random_ranges	6,780	7,663	1.13
covering_index_scan	7,695	7,257	0.94
groupby_scan	32,302	33,764	1.05
index_join	8,740	11,153	1.28
index_join_scan	5,276	5,892	1.12
types_table_scan	1,098,932	1,299,949	1.18
table_scan	1,303,664	1,417,268	1.09
oltp_read_only	160,483	172,069	1.07
Average			1.09

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	22,411	77,724	3.47
oltp_insert_ac	23,944	99,550	4.16
oltp_update_index_ac	26,798	106,429	3.97
oltp_update_non_index_ac	22,176	88,402	3.99
oltp_delete_insert_ac	23,561	99,380	4.22
oltp_write_only_ac	23,858	99,705	4.18
types_delete_insert_ac	21,816	86,756	3.98
oltp_read_write_ac	29,880	103,842	3.48
Average			3.93

100000 rows, median of 5 invocations per test; autocommit writes use 9, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (2.5x individual, 2x average; autocommit writes: 6x / 5x)

All tests within ceilings.

[github-actions]

Sysbench-Style Benchmark (TEXT PK): Doltlite vs SQLite

Companion to the classic Sysbench-Style Benchmark. Every workload here
runs against tables with a 32-char hex TEXT PRIMARY KEY (UUID-shaped).
Individual ratios gated at 2.5×; section averages gated at 2×. Autocommit writes use 6× / 5×.

In-Memory

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	23,502	28,321	1.21
oltp_range_select	11,407	11,536	1.01
oltp_sum_range	11,161	10,779	0.97
oltp_order_range	2,529	2,745	1.09
oltp_distinct_range	3,322	3,600	1.08
oltp_index_scan	3,690	4,947	1.34
select_random_points	15,583	18,036	1.16
select_random_ranges	3,183	4,542	1.43
covering_index_scan	3,555	3,548	1.00
groupby_scan	28,501	29,114	1.02
index_join	6,023	7,959	1.32
index_join_scan	4,101	4,925	1.20
types_table_scan	965,432	1,072,678	1.11
table_scan	1,353,687	1,244,116	0.92
oltp_read_only	103,311	113,041	1.09
Average			1.13

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	169,201	265,574	1.57
oltp_insert	17,070	30,577	1.79
oltp_update_index	59,361	109,759	1.85
oltp_update_non_index	39,583	66,786	1.69
oltp_delete_insert	42,657	87,590	2.05
oltp_write_only	23,976	50,140	2.09
types_delete_insert	26,206	45,309	1.73
oltp_read_write	72,239	113,548	1.57
Average			1.79

File-Backed

Reads

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	54,976	39,626	0.72
oltp_range_select	15,784	13,695	0.87
oltp_sum_range	15,194	12,726	0.84
oltp_order_range	3,159	3,096	0.98
oltp_distinct_range	4,255	4,050	0.95
oltp_index_scan	7,422	6,398	0.86
select_random_points	19,584	19,937	1.02
select_random_ranges	6,377	5,397	0.85
covering_index_scan	7,389	5,024	0.68
groupby_scan	31,181	29,430	0.94
index_join	8,364	9,574	1.14
index_join_scan	4,400	5,224	1.19
types_table_scan	910,591	1,020,369	1.12
table_scan	1,064,757	1,118,445	1.05
oltp_read_only	128,402	114,688	0.89
Average			0.94

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert	227,278	308,476	1.36
oltp_insert	56,492	54,394	0.96
oltp_update_index	205,934	198,074	0.96
oltp_update_non_index	151,269	126,736	0.84
oltp_delete_insert	166,196	173,145	1.04
oltp_write_only	121,234	118,894	0.98
types_delete_insert	128,307	120,773	0.94
oltp_read_write	173,885	170,214	0.98
Average			1.01

File-Backed (autocommit)

Each statement runs as its own transaction — exposes per-commit
fixed costs that the wrapped-in-BEGIN/COMMIT tests amortize away.
SQLite uses WAL mode with synchronous=FULL in this section so
the comparison uses SQLite's durable WAL autocommit path.

Reads

Reads have no commit cost; these are the same SQL files as the
File-Backed Reads section, included here for symmetry and to
catch any per-statement overhead doltlite pays on the read path.

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_point_select	31,121	35,588	1.14
oltp_range_select	12,626	13,321	1.06
oltp_sum_range	11,934	11,653	0.98
oltp_order_range	2,757	2,814	1.02
oltp_distinct_range	3,473	3,623	1.04
oltp_index_scan	5,039	5,884	1.17
select_random_points	16,293	18,430	1.13
select_random_ranges	4,241	5,078	1.20
covering_index_scan	5,172	4,637	0.90
groupby_scan	27,601	27,146	0.98
index_join	7,227	8,731	1.21
index_join_scan	4,235	5,095	1.20
types_table_scan	995,472	1,072,884	1.08
table_scan	1,149,951	1,222,685	1.06
oltp_read_only	111,917	123,547	1.10
Average			1.08

Writes

Test	SQLite (us)	Doltlite (us)	Multiplier
oltp_bulk_insert_ac	503,771	1,101,427	2.19
oltp_insert_ac	303,452	556,171	1.83
oltp_update_index_ac	209,610	522,458	2.49
oltp_update_non_index_ac	593,784	1,027,966	1.73
oltp_delete_insert_ac	292,521	252,183	0.86
oltp_write_only_ac	179,275	904,324	5.04
types_delete_insert_ac	537,252	558,701	1.04
oltp_read_write_ac	201,583	834,442	4.14
Average			2.42

100000 rows, median of 5 invocations per test; autocommit writes use 9, workload-only timing via host monotonic clock when available.

Performance Ceiling Check (2.5x individual, 2x average; autocommit writes: 6x / 5x)

All tests within ceilings.