vllm.v1.kv_offload.tiering.p2p.session.client ¶
Client-role state machine for a single peer session.
Handles outgoing fetch requests, abort-on-timeout, abort-ack timeout, and produces LoadResult for completed loads. The session coordinator parses wire messages and dispatches typed arguments here; this module never touches ControlConnection directly — it emits via the send callback injected by the coordinator (which gates on ConnectAck).
Classes:
-
ClientRole–Client-side load state machine for one peer session.
-
LoadResult–Result from a session poll, client side.
ClientRole ¶
Client-side load state machine for one peer session.
The coordinator owns the connection and the send-gating; this role is given a send callback and a peer_id for log messages and is otherwise self-contained.
Methods:
-
cancel–Cancel a pending load. Sends AbortFetchMsg if still active.
-
close–Tear down. Returns
(job_id, kv_request_id)for pending loads. -
collect_results–Walk timeouts and drain completed loads.
-
on_abort_ack–Handle an AbortAckMsg from the peer.
-
on_transfer_done–Handle a TransferDoneMsg from the peer.
-
request_blocks–Register a load request and send the FetchMsg.
Source code in vllm/v1/kv_offload/tiering/p2p/session/client.py
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 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 | |
cancel(kv_request_id) ¶
Cancel a pending load. Sends AbortFetchMsg if still active.
Source code in vllm/v1/kv_offload/tiering/p2p/session/client.py
close() ¶
Tear down. Returns (job_id, kv_request_id) for pending loads.
Source code in vllm/v1/kv_offload/tiering/p2p/session/client.py
collect_results() ¶
Walk timeouts and drain completed loads.
Active requests past _LOAD_TIMEOUT_S get an AbortFetchMsg sent and enter the aborting phase. Aborting requests past _ABORT_ACK_TIMEOUT_S are surfaced as failed loads.
Source code in vllm/v1/kv_offload/tiering/p2p/session/client.py
on_abort_ack(kv_request_id) ¶
Handle an AbortAckMsg from the peer.
Source code in vllm/v1/kv_offload/tiering/p2p/session/client.py
on_transfer_done(kv_request_id, success) ¶
Handle a TransferDoneMsg from the peer.
Source code in vllm/v1/kv_offload/tiering/p2p/session/client.py
request_blocks(job_id, kv_request_id, keys, block_ids, send_ready) ¶
Register a load request and send the FetchMsg.
Source code in vllm/v1/kv_offload/tiering/p2p/session/client.py
LoadResult ¶
_InboundRequestState dataclass ¶
Client-role state for a single load request.