vllm-project/vllm · error · ValueError

PCP does not support data parallelism yet.

Error message

PCP does not support data parallelism yet.

What it means

Prefill context parallelism (PCP) is not yet composable with data parallelism, so ParallelConfig raises when prefill_context_parallel_size > 1 and data_parallel_size > 1 simultaneously. This is a current implementation limitation, not a permanent rule.

Source

Thrown at vllm/config/parallel.py:528

                    "EPLB requires tensor, prefill-context, or data parallelism, "
                    f"but got TP={self.tensor_parallel_size}, "
                    f"PCP={self.prefill_context_parallel_size}, "
                    f"DP={self.data_parallel_size}."
                )
        else:
            if self.eplb_config.num_redundant_experts != 0:
                raise ValueError(
                    "num_redundant_experts is set to "
                    f"{self.eplb_config.num_redundant_experts} but EPLB is not "
                    "enabled. Either enable EPLB or unset "
                    "num_redundant_experts."
                )

        tp = self.tensor_parallel_size
        pcp = self.prefill_context_parallel_size
        dcp = self.decode_context_parallel_size
        if pcp > 1 and self.data_parallel_size > 1:
            raise ValueError("PCP does not support data parallelism yet.")
        if pcp == 1:
            # DCP reuses the TP ranks when PCP is disabled.
            if tp % dcp != 0:
                raise ValueError(f"tp_size={tp} must be divisible by dcp_size={dcp}.")
        elif dcp not in (1, pcp, tp * pcp):
            raise ValueError(
                "When PCP is enabled, DCP must be disabled, span the PCP "
                "axis, or span the full TP x PCP axis. "
                f"Got TP={tp}, PCP={pcp}, DCP={dcp}; valid DCP sizes are "
                f"{sorted({1, pcp, tp * pcp})}."
            )

        if self.dcp_comm_backend == "a2a" and self.decode_context_parallel_size <= 1:
            raise ValueError(
                "dcp_comm_backend='a2a' requires decode_context_parallel_size > 1."
            )

        return self

View on GitHub (pinned to c794754062)

Solutions

1. Drop data parallelism (set --data-parallel-size 1) and keep PCP.
2. Or drop PCP (set prefill_context_parallel_size back to 1) and keep DP.
3. Track vLLM release notes; re-test the combination once PCP+DP support lands.

Example fix

# before
vllm serve model --prefill-context-parallel-size 2 --data-parallel-size 2
# after
vllm serve model --prefill-context-parallel-size 2 --data-parallel-size 1

def pcp_dp_valid(pcp: int, dp: int) -> bool:
    return pcp <= 1 or dp <= 1

assert pcp_dp_valid(1, 4)

When it happens

Trigger: Launching with both --prefill-context-parallel-size > 1 (e.g. via --enable-prefix-caching style PCP flags) and --data-parallel-size > 1.

Common situations: Long-context deployments adding DP for throughput on top of an existing PCP config; upgrading vLLM and combining previously separate feature flag sets.

Related errors

• data_parallel_size_local ({self.data_parallel_size_local}) m
• data_parallel_external_lb can only be set when data_parallel
• tp_size={tp} must be divisible by dcp_size={dcp}.
• When PCP is enabled, DCP must be disabled, span the PCP axis
• data parallel rank {rank} is not connected to this frontend;

AI-assisted analysis of vllm-project/vllm@c794754062 (2026-08-14). Data as JSON: /api/errors/d172e4fffc856c3a. Report an issue: GitHub.