zzstoatzz.io
atproto relay implementation in zig
zlay.waow.tech
feat: multi-threaded DID resolver, cache warming, and listHosts endpoint
spawn N resolver threads (default 4, configurable via RESOLVER_THREADS)
instead of 1, and pre-populate the resolve queue with all active DIDs
from the database on startup. adds com.atproto.sync.listHosts XRPC
endpoint with Indigo-compatible response format (includes seq field,
filters by last_seq > 0, default limit 200).
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
+76
src/main.zig
+76
src/main.zig
···
11
11
//! /xrpc/com.atproto.sync.getRepoStatus — single account status
12
12
//! /xrpc/com.atproto.sync.getLatestCommit — latest commit CID + rev
13
13
//! /xrpc/com.atproto.sync.listReposByCollection — repos with records in a collection
14
+
//! /xrpc/com.atproto.sync.listHosts — paginated active host listing
14
15
//! /xrpc/com.atproto.sync.requestCrawl — request PDS crawl (POST)
15
16
//! /admin/hosts — list all hosts (GET, admin)
16
17
//! /admin/hosts/block — block a host (POST, admin)
···
117
118
118
119
// start: loads active hosts from DB, spawns subscriber threads
119
120
try slurper.start();
121
+
122
+
// pre-populate resolver queue with all known active DIDs
123
+
val.warmCache(&dp);
120
124
121
125
// start GC thread (runs every 10 minutes)
122
126
const gc_thread = try std.Thread.spawn(.{}, gcLoop, .{&dp});
···
259
263
handleGetLatestCommit(request, query, persist);
260
264
} else if (std.mem.eql(u8, path, "/xrpc/com.atproto.sync.listReposByCollection")) {
261
265
handleListReposByCollection(request, query, ci);
266
+
} else if (std.mem.eql(u8, path, "/xrpc/com.atproto.sync.listHosts")) {
267
+
handleListHosts(request, query, persist);
262
268
} else if (std.mem.eql(u8, path, "/admin/hosts")) {
263
269
handleAdminListHosts(request, persist, slurper);
264
270
} else if (std.mem.eql(u8, path, "/")) {
···
802
808
w.writeAll(last) catch return;
803
809
w.writeAll("\"") catch return;
804
810
}
811
+
}
812
+
813
+
w.writeByte('}') catch return;
814
+
respondJson(request, .ok, fbs.getWritten());
815
+
}
816
+
817
+
fn handleListHosts(request: *http.Server.Request, query: []const u8, persist: *event_log_mod.DiskPersist) void {
818
+
const cursor_str = queryParam(query, "cursor") orelse "0";
819
+
const limit_str = queryParam(query, "limit") orelse "200";
820
+
821
+
const cursor_val = std.fmt.parseInt(i64, cursor_str, 10) catch {
822
+
respondJson(request, .bad_request, "{\"error\":\"BadRequest\",\"message\":\"invalid cursor\"}");
823
+
return;
824
+
};
825
+
if (cursor_val < 0) {
826
+
respondJson(request, .bad_request, "{\"error\":\"BadRequest\",\"message\":\"cursor must be >= 0\"}");
827
+
return;
828
+
}
829
+
830
+
const limit = std.fmt.parseInt(i64, limit_str, 10) catch {
831
+
respondJson(request, .bad_request, "{\"error\":\"BadRequest\",\"message\":\"invalid limit\"}");
832
+
return;
833
+
};
834
+
if (limit < 1 or limit > 1000) {
835
+
respondJson(request, .bad_request, "{\"error\":\"BadRequest\",\"message\":\"limit must be 1..1000\"}");
836
+
return;
837
+
}
838
+
839
+
var result = persist.db.query(
840
+
"SELECT id, hostname, status, last_seq FROM host WHERE id > $1 AND last_seq > 0 ORDER BY id ASC LIMIT $2",
841
+
.{ cursor_val, limit },
842
+
) catch {
843
+
respondJson(request, .internal_server_error, "{\"error\":\"DatabaseError\",\"message\":\"query failed\"}");
844
+
return;
845
+
};
846
+
defer result.deinit();
847
+
848
+
var buf: [65536]u8 = undefined;
849
+
var fbs = std.io.fixedBufferStream(&buf);
850
+
const w = fbs.writer();
851
+
852
+
var count: i64 = 0;
853
+
var last_id: i64 = 0;
854
+
855
+
w.writeAll("{\"hosts\":[") catch return;
856
+
857
+
while (result.nextUnsafe() catch null) |row| {
858
+
if (count > 0) w.writeByte(',') catch return;
859
+
860
+
const id = row.get(i64, 0);
861
+
const hostname = row.get([]const u8, 1);
862
+
const status = row.get([]const u8, 2);
863
+
const seq = row.get(i64, 3);
864
+
865
+
w.writeAll("{\"hostname\":\"") catch return;
866
+
w.writeAll(hostname) catch return;
867
+
w.writeAll("\"") catch return;
868
+
std.fmt.format(w, ",\"seq\":{d}", .{seq}) catch return;
869
+
w.writeAll(",\"status\":\"") catch return;
870
+
w.writeAll(status) catch return;
871
+
w.writeAll("\"}") catch return;
872
+
873
+
last_id = id;
874
+
count += 1;
875
+
}
876
+
877
+
w.writeByte(']') catch return;
878
+
879
+
if (count >= limit and count > 1) {
880
+
std.fmt.format(w, ",\"cursor\":\"{d}\"", .{last_id}) catch return;
805
881
}
806
882
807
883
w.writeByte('}') catch return;
+44
-5
src/validator.zig
+44
-5
src/validator.zig
···
8
8
const std = @import("std");
9
9
const zat = @import("zat");
10
10
const broadcaster = @import("broadcaster.zig");
11
+
const event_log_mod = @import("event_log.zig");
11
12
12
13
const Allocator = std.mem.Allocator;
13
14
const log = std.log.scoped(.relay);
···
50
51
queue: std.ArrayListUnmanaged([]const u8) = .{},
51
52
queue_mutex: std.Thread.Mutex = .{},
52
53
queue_cond: std.Thread.Condition = .{},
53
-
resolver_thread: ?std.Thread = null,
54
+
resolver_threads: [max_resolver_threads]?std.Thread = .{null} ** max_resolver_threads,
54
55
alive: std.atomic.Value(bool) = .{ .raw = true },
56
+
57
+
const max_resolver_threads = 8;
58
+
const default_resolver_threads = 4;
55
59
56
60
pub fn init(allocator: Allocator, stats: *broadcaster.Stats) Validator {
57
61
return initWithConfig(allocator, stats, .{});
···
67
71
68
72
pub fn deinit(self: *Validator) void {
69
73
self.alive.store(false, .release);
70
-
self.queue_cond.signal();
71
-
if (self.resolver_thread) |t| t.join();
74
+
self.queue_cond.broadcast();
75
+
for (&self.resolver_threads) |*t| {
76
+
if (t.*) |thread| {
77
+
thread.join();
78
+
t.* = null;
79
+
}
80
+
}
72
81
73
82
// free cache keys (CachedKey is inline, no separate free needed)
74
83
var cache_it = self.cache.iterator();
···
84
93
self.queue.deinit(self.allocator);
85
94
}
86
95
87
-
/// start the background resolver thread
96
+
/// start background resolver threads
88
97
pub fn start(self: *Validator) !void {
89
-
self.resolver_thread = try std.Thread.spawn(.{}, resolveLoop, .{self});
98
+
const n = parseEnvInt(u8, "RESOLVER_THREADS", default_resolver_threads);
99
+
const count = @min(n, max_resolver_threads);
100
+
for (self.resolver_threads[0..count]) |*t| {
101
+
t.* = try std.Thread.spawn(.{}, resolveLoop, .{self});
102
+
}
103
+
}
104
+
105
+
/// pre-populate the resolve queue with all active DIDs from the database.
106
+
/// call after slurper.start() so subscribers are already running.
107
+
pub fn warmCache(self: *Validator, persist: *event_log_mod.DiskPersist) void {
108
+
var result = persist.db.query(
109
+
"SELECT did FROM account WHERE status = 'active' AND upstream_status = 'active' ORDER BY uid ASC",
110
+
.{},
111
+
) catch |err| {
112
+
log.warn("cache warming: query failed: {s}", .{@errorName(err)});
113
+
return;
114
+
};
115
+
defer result.deinit();
116
+
117
+
var count: u64 = 0;
118
+
while (result.nextUnsafe() catch null) |row| {
119
+
const did = row.get([]const u8, 0);
120
+
self.queueResolve(did);
121
+
count += 1;
122
+
}
123
+
log.info("cache warming: queued {d} DIDs for resolution", .{count});
90
124
}
91
125
92
126
/// validate a #sync frame: signature verification only (no ops, no MST).
···
443
477
return self.cache.count();
444
478
}
445
479
};
480
+
481
+
fn parseEnvInt(comptime T: type, key: []const u8, default: T) T {
482
+
const val = std.posix.getenv(key) orelse return default;
483
+
return std.fmt.parseInt(T, val, 10) catch default;
484
+
}
446
485
447
486
// --- tests ---
448
487