test_service.py 8.81 KB
Newer Older
one's avatar
one committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
"""Tests for hytop.gpu.service — business logic (collect_from_host mocked)."""

from __future__ import annotations

import json
import time
from unittest.mock import patch

from hytop.core.history import SlidingHistory
from hytop.core.ssh import CollectResult
from hytop.gpu.models import MonitorState, NodeResult, Sample
from hytop.gpu.service import (
    apply_node_results,
    availability_ready,
    collect_node,
    drain_pending_nodes,
    init_monitor_state,
)

# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------

HY_SMI_FULL_JSON = json.dumps(
    {
        "card0": {
            "Average Graphics Package Power (W)": "157.0",
            "Temperature (Sensor core) (C)": "30.0",
            "HCU use (%)": "0.0",
            "HCU memory use (%)": "89",
            "sclk clock speed": "1500Mhz",
        },
        "card7": {
            "Average Graphics Package Power (W)": "141.0",
            "Temperature (Sensor core) (C)": "25.0",
            "HCU use (%)": "100.0",
            "HCU memory use (%)": "89",
            "sclk clock speed": "1500Mhz",
        },
    }
)


def _state(hosts=("localhost",), device_filter=None, max_window=10.0) -> MonitorState:
    return init_monitor_state(hosts=list(hosts), device_filter=device_filter, max_window=max_window)


one's avatar
one committed
48
49
def _sample(ts: float, gpu_pct: float = 0.0, vram_pct: float = 0.0) -> Sample:
    return Sample(ts=ts, gpu_pct=gpu_pct, vram_pct=vram_pct)
one's avatar
one committed
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79


# ---------------------------------------------------------------------------
# collect_node
# ---------------------------------------------------------------------------


class TestCollectNode:
    @patch("hytop.gpu.service.collect_from_host")
    def test_success_returns_samples(self, mock_collect):
        mock_collect.return_value = CollectResult(
            host="localhost", stdout=HY_SMI_FULL_JSON, stderr=""
        )
        result = collect_node("localhost", ssh_timeout=5, cmd_timeout=10, hy_smi_args=["--json"])
        assert result.error is None
        assert set(result.samples.keys()) == {0, 7}

    @patch("hytop.gpu.service.collect_from_host")
    def test_host_error_propagated(self, mock_collect):
        mock_collect.return_value = CollectResult(
            host="node01", stdout="", stderr="", error="timeout after 10.0s"
        )
        result = collect_node("node01", ssh_timeout=5, cmd_timeout=10, hy_smi_args=["--json"])
        assert result.error is not None
        assert "timeout" in result.error

    @patch("hytop.gpu.service.collect_from_host")
    def test_empty_output_yields_error(self, mock_collect):
        mock_collect.return_value = CollectResult(host="localhost", stdout="", stderr="")
        result = collect_node("localhost", ssh_timeout=5, cmd_timeout=10, hy_smi_args=["--json"])
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
        assert result.error == "no gpu rows parsed (empty output)"

    @patch("hytop.gpu.service.collect_from_host")
    def test_noise_before_valid_json_is_parsed(self, mock_collect):
        noisy_stdout = (
            "Authorized users only.\\n"
            "RTNETLINK answers: File exists\\n"
            + HY_SMI_FULL_JSON
        )
        mock_collect.return_value = CollectResult(host="localhost", stdout=noisy_stdout, stderr="")
        result = collect_node("localhost", ssh_timeout=5, cmd_timeout=10, hy_smi_args=["--json"])
        assert result.error is None
        assert set(result.samples.keys()) == {0, 7}

    @patch("hytop.gpu.service.collect_from_host")
    def test_invalid_json_yields_specific_error(self, mock_collect):
        mock_collect.return_value = CollectResult(host="localhost", stdout="not json", stderr="")
        result = collect_node("localhost", ssh_timeout=5, cmd_timeout=10, hy_smi_args=["--json"])
        assert result.error == "no gpu rows parsed (invalid json output)"
one's avatar
one committed
99
100
101
102
103
104
105
106
107
108


# ---------------------------------------------------------------------------
# apply_node_results
# ---------------------------------------------------------------------------


class TestApplyNodeResults:
    def test_successful_node_adds_to_history(self):
        state = _state(hosts=["localhost"])
one's avatar
one committed
109
        sample = _sample(ts=1.0, gpu_pct=50.0)
one's avatar
one committed
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
        node = NodeResult(host="localhost", samples={0: sample})
        apply_node_results([node], device_filter=None, state=state)
        assert ("localhost", 0) in state.histories

    def test_error_node_sets_error(self):
        state = _state(hosts=["localhost"])
        node = NodeResult(host="localhost", samples={}, error="connection refused")
        apply_node_results([node], device_filter=None, state=state)
        assert state.errors["localhost"] == "connection refused"

    def test_device_filter_excludes_other_gpus(self):
        state = _state(hosts=["localhost"], device_filter={0})
        s0 = _sample(ts=1.0)
        s1 = _sample(ts=1.0)
        node = NodeResult(host="localhost", samples={0: s0, 1: s1})
        apply_node_results([node], device_filter={0}, state=state)
        assert ("localhost", 0) in state.histories
        assert ("localhost", 1) not in state.histories

    def test_success_clears_previous_error(self):
        state = _state(hosts=["localhost"])
        state.errors["localhost"] = "old error"
        node = NodeResult(host="localhost", samples={0: _sample(ts=1.0)})
        apply_node_results([node], device_filter=None, state=state)
        assert "localhost" not in state.errors

    def test_duplicate_sample_not_added(self):
        state = _state(hosts=["localhost"])
        sample = _sample(ts=5.0)
        node = NodeResult(host="localhost", samples={0: sample})
        apply_node_results([node], device_filter=None, state=state)
        apply_node_results([node], device_filter=None, state=state)  # same ts
        assert len(state.histories[("localhost", 0)].samples) == 1


# ---------------------------------------------------------------------------
# availability_ready
# ---------------------------------------------------------------------------


class TestAvailabilityReady:
one's avatar
one committed
151
    def _make_history(self, gpu_pct: float, vram_pct: float) -> SlidingHistory:
one's avatar
one committed
152
153
        """Build a SlidingHistory with one fresh sample using real monotonic time."""
        h = SlidingHistory(max_window_s=30)
one's avatar
one committed
154
        h.add(_sample(ts=time.monotonic(), gpu_pct=gpu_pct, vram_pct=vram_pct))
one's avatar
one committed
155
156
157
158
        return h

    def test_idle_gpu_returns_true(self):
        key = ("localhost", 0)
one's avatar
one committed
159
        histories = {key: self._make_history(gpu_pct=0.0, vram_pct=0.0)}
one's avatar
one committed
160
161
162
163
164
165
166
167
168
169
        assert availability_ready(
            window=5.0,
            histories=histories,
            monitored_keys={key},
            hosts=["localhost"],
            errors={},
        )

    def test_busy_gpu_returns_false(self):
        key = ("localhost", 0)
one's avatar
one committed
170
        histories = {key: self._make_history(gpu_pct=100.0, vram_pct=89.0)}
one's avatar
one committed
171
172
173
174
175
176
177
178
179
180
        assert not availability_ready(
            window=5.0,
            histories=histories,
            monitored_keys={key},
            hosts=["localhost"],
            errors={},
        )

    def test_host_error_returns_false(self):
        key = ("localhost", 0)
one's avatar
one committed
181
        histories = {key: self._make_history(gpu_pct=0.0, vram_pct=0.0)}
one's avatar
one committed
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
        assert not availability_ready(
            window=5.0,
            histories=histories,
            monitored_keys={key},
            hosts=["localhost"],
            errors={"localhost": "connection refused"},
        )

    def test_empty_monitored_keys_returns_false(self):
        assert not availability_ready(
            window=5.0,
            histories={},
            monitored_keys=set(),
            hosts=["localhost"],
            errors={},
        )

    def test_missing_history_returns_false(self):
        key = ("localhost", 0)
        assert not availability_ready(
            window=5.0,
            histories={},  # no history for this key
            monitored_keys={key},
            hosts=["localhost"],
            errors={},
        )


# ---------------------------------------------------------------------------
# drain_pending_nodes
# ---------------------------------------------------------------------------


class TestDrainPendingNodes:
    def test_drains_new_result(self):
        state = _state(hosts=["localhost"])
        result = NodeResult(host="localhost", samples={})
        with state.state_lock:
            snap = state.host_state["localhost"]
            snap.seq = 1
            snap.result = result

        nodes = drain_pending_nodes(["localhost"], state)
        assert len(nodes) == 1
        assert nodes[0] is result

    def test_does_not_drain_already_processed(self):
        state = _state(hosts=["localhost"])
        state.processed_seq["localhost"] = 1
        with state.state_lock:
            snap = state.host_state["localhost"]
            snap.seq = 1
            snap.result = NodeResult(host="localhost", samples={})

        nodes = drain_pending_nodes(["localhost"], state)
        assert nodes == []