upstream: an unnamed connect errno fails over, local-resource exchanges are counted, doh faults carry the http status

src/upstream/doh_client.zig

@@ -123,7 +123,11 @@ pub const DohClient = struct {
         // is an error before the location is ever read.
         var resp = req.receiveHead(&.{}) catch |err| return fault(headCause(&req, err), .receive);
 
-        if (resp.head.status != .ok) return peerFault(error.HttpStatus, error.HttpStatus);
+        if (resp.head.status != .ok) return .{ .fault = .{
+            .kind = error.HttpStatus,
+            .cause = error.HttpStatus,
+            .http_status = resp.head.status,
+        } };
         // `head.content_type` points into memory that `resp.reader` invalidates,
         // so the check happens before the body stream starts.
         if (!contentTypeOk(resp.head.content_type)) return peerFault(error.HttpContentType, error.HttpContentType);
@@ -175,7 +179,19 @@ const Phase = enum { connect, send, receive };
 ///   arrive here as themselves. Without them a decode error or a bad record MAC
 ///   would be reported as a plain receive failure.
 fn fault(err: anyerror, phase: Phase) transport.LeafError!transport.Outcome {
-    if (transport.mapLocal(err)) |local| return local;
+    // `.connect` is the whole of `http.request`: the name lookup, the TCP
+    // connect and the handshake. The handshake's only `error.Unexpected` is the
+    // errno the socket stream stored (std/http/Client.zig:1466; the TLS
+    // client's `InitError` has none), and a CA bundle scan fails as
+    // `CertificateBundleLoadFailure`. The lookup's resolver socket can fail
+    // locally, but that still leaves this upstream unreachable while a DoT
+    // upstream, always an IP literal, needs no lookup, so the fault belongs to
+    // this endpoint and the pool is right to move on.
+    const local = switch (phase) {
+        .connect => transport.mapLocalConnect(err),
+        .send, .receive => transport.mapLocal(err),
+    };
+    if (local) |claimed| return claimed;
     return peerFault(kindOf(err, phase), err);
 }
 
@@ -331,6 +347,12 @@ test "a local cause stays an error instead of becoming a fault" {
     try testing.expectError(error.Unexpected, fault(error.Unexpected, .send));
 }
 
+test "an unnamed connect errno is a connect fault, so the pool fails over" {
+    const outcome = try fault(error.Unexpected, .connect);
+    try testing.expectEqual(transport.PeerFault.ConnectFailed, outcome.fault.kind);
+    try testing.expectEqual(error.Unexpected, outcome.fault.cause);
+}
+
 test "kindOf maps the collapsed tls errors regardless of phase" {
     try testing.expectEqual(error.TlsFailed, kindOf(error.TlsInitializationFailed, .connect));
     try testing.expectEqual(error.TlsFailed, kindOf(error.TlsInitializationFailed, .receive));

src/upstream/dot_client.zig

@@ -279,7 +279,8 @@ pub const DotClient = struct {
 
         const stream = address.connect(io, .{ .mode = .stream }) catch |err| {
             log.debug("{f}", .{self.diagnose(.{ .connect_failed = err })});
-            return try transport.faultOrLocal(err, error.ConnectFailed);
+            if (transport.mapLocalConnect(err)) |local| return local;
+            return .{ .kind = error.ConnectFailed, .cause = err };
         };
 
         // Emplaced before the handshake, never built beside it and copied in:

src/upstream/health.zig

@@ -56,10 +56,10 @@ comptime {
 pub const window_len = 32;
 
 /// Bytes kept of a rendered `transport.Fault`, truncated to fit. Sized so the
-/// widest classification and the widest concrete cause name fit whole; the
-/// comptime-driven test below recomputes that worst case from the std error
-/// sets, so a wider name std adds fails the test run rather than silently
-/// truncating an operator's only diagnostic.
+/// widest classification and the widest concrete cause name fit whole, and so
+/// does the widest HTTP status with its phrase; the comptime-driven test below
+/// recomputes that worst case from std, so a wider name std adds fails the test
+/// run rather than silently truncating an operator's only diagnostic.
 pub const error_name_capacity = 64;
 
 /// The shift is capped so `base_backoff_ms << shift` cannot run away; by then
@@ -552,7 +552,7 @@ test "lastError truncates a cause too wide for the buffer instead of overflowing
     try testing.expectEqualStrings(whole[0..error_name_capacity], state.lastError());
 }
 
-test "error_name_capacity holds the widest classification and cause whole" {
+test "error_name_capacity holds the widest classification, cause and status whole" {
     // The widest concrete cause an upstream unwrap can produce. The three sets
     // are the ones `transport.zig`'s unwraps return a member of; a wider name
     // std adds fails here rather than truncating an operator's diagnostic.
@@ -579,6 +579,18 @@ test "error_name_capacity holds the widest classification and cause whole" {
 
     try testing.expectEqualStrings("DetectingNetworkConfigurationFailed", widest_cause);
     try testing.expect(widest_kind.len + " (cause ".len + widest_cause.len + ")".len <= error_name_capacity);
+
+    const widest_phrase = comptime blk: {
+        var widest: []const u8 = "";
+        for (@typeInfo(std.http.Status).@"enum".fields) |field| {
+            const phrase = @as(std.http.Status, @enumFromInt(field.value)).phrase() orelse continue;
+            if (phrase.len > widest.len) widest = phrase;
+        }
+        break :blk widest;
+    };
+    // `std.http.Status` is a u10, so a code has at most four digits.
+    const widest_status = "HttpStatus (status ".len + 4 + " ".len + widest_phrase.len + ")".len;
+    try testing.expect(widest_status <= error_name_capacity);
 }
 
 /// The fault of a `tripped` effect, or an error when the effect says `clear`.

src/upstream/pool.zig

@@ -382,6 +382,13 @@ pub const Pool = struct {
     /// removes. Read through `budgetExhaustedTotal`, since `Snapshot` is
     /// per-entry and has nowhere to put it.
     budget_exhausted_total: std.atomic.Value(u64) = .init(0),
+    /// Exchanges this process ended on its own resources (memory, descriptors,
+    /// an undersized buffer, an errno std does not name outside the connect
+    /// phase), counted once each by `exchange`. Pool-wide for the reason
+    /// `budget_exhausted_total` is: no endpoint caused it, so the pool records
+    /// it against none, and the SERVFAIL it becomes names no upstream. Read
+    /// through `localResourceTotal`.
+    local_resource_total: std.atomic.Value(u64) = .init(0),
 
     /// The failover loop tracks which entries it has already spent in a bitset
     /// on its stack, so the number of upstreams one pool may hold is bounded.
@@ -461,11 +468,13 @@ pub const Pool = struct {
     ) transport.ExchangeError![]u8 {
         const deadline: std.Io.Clock.Timestamp = .fromNow(io, self.timeouts.total);
         return self.failover(io, query, response_buf, selected, deadline) catch |err| {
-            // Once per exhausted exchange, wherever the exhaustion was
-            // detected — the loop's own waits and attempts, or a leaf that
-            // reported one of its own.
-            if (err == error.BudgetExhausted) {
-                _ = self.budget_exhausted_total.fetchAdd(1, .monotonic);
+            // Once per failed exchange, wherever the failure was detected —
+            // the loop's own waits and attempts, or a leaf that reported one of
+            // its own.
+            switch (transport.group(err)) {
+                .budget_exhausted => _ = self.budget_exhausted_total.fetchAdd(1, .monotonic),
+                .local_resource => _ = self.local_resource_total.fetchAdd(1, .monotonic),
+                .peer_fault, .cancellation => {},
             }
             return err;
         };
@@ -518,6 +527,11 @@ pub const Pool = struct {
         return self.budget_exhausted_total.load(.acquire);
     }
 
+    /// Exchanges that ended on this process's own resources, pool-wide.
+    pub fn localResourceTotal(self: *const Pool) u64 {
+        return self.local_resource_total.load(.acquire);
+    }
+
     /// The two-pass failover loop.
     ///
     /// Pass one sweeps the entries health says are available, in priority
@@ -1431,6 +1445,11 @@ test "a local resource error short-circuits and records nothing" {
     try testing.expectEqual(@as(usize, 0), good.calls.load(.acquire));
     try testing.expectEqual(@as(u64, 0), entries[0].health.total_failures);
     try testing.expectEqual(@as(u32, 0), entries[0].health.consecutive_failures);
+    // Recorded against no endpoint, but still counted: the SERVFAIL it becomes
+    // names no upstream, and this is the only trace of why.
+    try testing.expectEqual(@as(u64, 1), pool.localResourceTotal());
+    try testing.expectEqual(@as(u64, 0), pool.budgetExhaustedTotal());
+    try testing.expect(selected == null);
 }
 
 test "a cancellation short-circuits and records nothing" {
@@ -1453,6 +1472,7 @@ test "a cancellation short-circuits and records nothing" {
     try testing.expectEqual(@as(u64, 0), entries[0].health.total_failures);
     // A torn-down task is not a budget that ran out, and it names no endpoint.
     try testing.expectEqual(@as(u64, 0), pool.budgetExhaustedTotal());
+    try testing.expectEqual(@as(u64, 0), pool.localResourceTotal());
     try testing.expect(selected == null);
 }
 
@@ -1722,6 +1742,7 @@ test "budget_exhausted_total counts exhausted exchanges once each" {
     stalling.behavior = .{ .reply = response_bytes };
     _ = try pool.exchange(io, query_bytes, &buf, &selected);
     try testing.expectEqual(@as(u64, 3), pool.budgetExhaustedTotal());
+    try testing.expectEqual(@as(u64, 0), pool.localResourceTotal());
 }
 
 test "two stalling upstreams cost the total budget, not one budget each" {

src/upstream/transport.zig

@@ -263,6 +263,20 @@ pub fn mapPhase(err: anyerror, phase: PeerFault) ExchangeError {
     return mapLocal(err) orelse phase;
 }
 
+/// `mapLocal` for the connect phase, where `error.Unexpected` belongs to the
+/// peer.
+///
+/// zig 0.16.0's `connect()` names no `EINVAL` (std/Io/Threaded.zig:11909), and
+/// `EINVAL` is what a blackhole route to the peer returns, so a route fault
+/// arrives as `error.Unexpected`. Claimed as local, it would end the exchange
+/// without failover and record nothing against the endpoint. In this phase an
+/// errno std does not name is about the path to this peer, so it stays a peer
+/// fault and the pool moves on to the next endpoint.
+pub fn mapLocalConnect(err: anyerror) ?LeafError {
+    if (err == error.Unexpected) return null;
+    return mapLocal(err);
+}
+
 /// Closes `target` with cancellation blocked.
 ///
 /// A transport's close runs from a `defer` chain that a lost timeout race is
@@ -458,10 +472,19 @@ pub const Client = struct {
 pub const Fault = struct {
     kind: PeerFault,
     cause: anyerror,
+    /// The status the peer answered with, when that is the fault. An error
+    /// name cannot carry it, and it is what says why the peer refused: a 505
+    /// is a DoH server that will not speak HTTP/1.1.
+    http_status: ?std.http.Status = null,
 
-    /// The one text every surface prints. `<Kind> (cause <Cause>)`.
+    /// The one text every surface prints: `<Kind> (cause <Cause>)`, or
+    /// `<Kind> (status <code> <phrase>)` for a peer that answered with a status.
     pub fn format(self: Fault, w: *std.Io.Writer) std.Io.Writer.Error!void {
-        try w.print("{t} (cause {s})", .{ self.kind, @errorName(self.cause) });
+        const status = self.http_status orelse
+            return w.print("{t} (cause {s})", .{ self.kind, @errorName(self.cause) });
+        try w.print("{t} (status {d}", .{ self.kind, @intFromEnum(status) });
+        if (status.phrase()) |phrase| try w.print(" {s}", .{phrase});
+        try w.writeByte(')');
     }
 };
 
@@ -776,6 +799,37 @@ test "mapPhase names the phase unless the error is this process's own" {
     );
 }
 
+test "mapLocalConnect leaves an unnamed connect errno to the peer" {
+    try testing.expectEqual(@as(?ExchangeError, null), mapLocalConnect(error.Unexpected));
+    try testing.expectEqual(@as(?ExchangeError, error.OutOfMemory), mapLocalConnect(error.OutOfMemory));
+    try testing.expectEqual(
+        @as(?ExchangeError, error.ProcessFdQuotaExceeded),
+        mapLocalConnect(error.ProcessFdQuotaExceeded),
+    );
+    try testing.expectEqual(@as(?ExchangeError, error.Canceled), mapLocalConnect(error.Canceled));
+    try testing.expectEqual(@as(?ExchangeError, null), mapLocalConnect(error.NetworkUnreachable));
+}
+
+test "a fault with a status prints the code and its phrase" {
+    var buf: [64]u8 = undefined;
+    const refused: Fault = .{
+        .kind = error.HttpStatus,
+        .cause = error.HttpStatus,
+        .http_status = .http_version_not_supported,
+    };
+    try testing.expectEqualStrings(
+        "HttpStatus (status 505 HTTP Version Not Supported)",
+        try std.fmt.bufPrint(&buf, "{f}", .{refused}),
+    );
+
+    // A code std has no name for still prints, without a phrase.
+    const unnamed: Fault = .{ .kind = error.HttpStatus, .cause = error.HttpStatus, .http_status = @enumFromInt(599) };
+    try testing.expectEqualStrings("HttpStatus (status 599)", try std.fmt.bufPrint(&buf, "{f}", .{unnamed}));
+
+    const plain: Fault = .{ .kind = error.ConnectFailed, .cause = error.Unexpected };
+    try testing.expectEqualStrings("ConnectFailed (cause Unexpected)", try std.fmt.bufPrint(&buf, "{f}", .{plain}));
+}
+
 /// Stands in for whatever the pool or the forward client races. The variants
 /// are the three ways such a task ends: in time, too late, or with a failure of
 /// its own.

src/web/metrics.zig

@@ -186,6 +186,9 @@ pub const Sample = struct {
     /// label and cannot live in `UpstreamSample`. Absent while no pool is
     /// wired, like every other collaborator.
     upstream_budget_exhausted_total: ?u64 = null,
+    /// Exchanges the pool ended on this process's own resources. Pool-wide for
+    /// the same reason as `upstream_budget_exhausted_total`.
+    upstream_local_resource_total: ?u64 = null,
     peaks: PeakSample = .{},
 };
 
@@ -300,6 +303,7 @@ pub fn collect(state: *server.WebState, io: std.Io, arena: Allocator) Allocator.
         if (generation.pool) |pool| {
             sample.upstreams = try upstreams(pool, io, arena);
             sample.upstream_budget_exhausted_total = pool.budgetExhaustedTotal();
+            sample.upstream_local_resource_total = pool.localResourceTotal();
         }
     }
 
@@ -512,6 +516,15 @@ pub fn render(w: *std.Io.Writer, sample: Sample) std.Io.Writer.Error!void {
         );
         try w.print("nxdns_upstream_budget_exhausted_total {d}\n", .{total});
     }
+    if (sample.upstream_local_resource_total) |total| {
+        try labeledHead(
+            w,
+            "nxdns_upstream_local_resource_total",
+            "Exchanges this process ended on its own resources (memory, descriptors, an unnamed system error) before any upstream answered.",
+            "counter",
+        );
+        try w.print("nxdns_upstream_local_resource_total {d}\n", .{total});
+    }
 }
 
 fn renderPeaks(w: *std.Io.Writer, peaks: PeakSample) std.Io.Writer.Error!void {
@@ -1788,3 +1801,20 @@ test "the budget-exhausted counter renders pool-wide, without a url label" {
     defer testing.allocator.free(absent);
     try testing.expect(!std.mem.containsAtLeast(u8, absent, 1, "nxdns_upstream_budget_exhausted_total"));
 }
+
+test "the local-resource counter renders pool-wide, without a url label" {
+    const text = try renderToString(testing.allocator, .{ .upstream_local_resource_total = 4 });
+    defer testing.allocator.free(text);
+
+    try testing.expect(std.mem.containsAtLeast(
+        u8,
+        text,
+        1,
+        "# TYPE nxdns_upstream_local_resource_total counter\n",
+    ));
+    try testing.expect(std.mem.containsAtLeast(u8, text, 1, "nxdns_upstream_local_resource_total 4\n"));
+
+    const absent = try renderToString(testing.allocator, .{});
+    defer testing.allocator.free(absent);
+    try testing.expect(!std.mem.containsAtLeast(u8, absent, 1, "nxdns_upstream_local_resource_total"));
+}