From d96af2688bf1002660b054c098583fac03d12e4b Mon Sep 17 00:00:00 2001 From: Agata Dobrzyniewicz Date: Fri, 21 Nov 2025 08:32:41 +0200 Subject: [PATCH 1/4] Troubleshooting - async + preemption Signed-off-by: Agata Dobrzyniewicz --- docs/user_guide/troubleshooting.md | 12 ++++++++++++ 1 file changed, 12 insertions(+) diff --git a/docs/user_guide/troubleshooting.md b/docs/user_guide/troubleshooting.md index c26d7ad6ad..3e907f1308 100644 --- a/docs/user_guide/troubleshooting.md +++ b/docs/user_guide/troubleshooting.md @@ -27,3 +27,15 @@ Vllm calculates maximum available concurrency for current environment based on k ``` So for this specific scenario correct value for --max_concurrency is 10 + +### 3. Async scheduler causes error after preemption: + +``` +(EngineCore_DP0 pid=19906) assert request.num_output_placeholders >= 0 +``` + +#### Solution: +0.11.0 version is missing crucial fixes from upstream. To avoid it try: +- lowering max conccurency +- decresing `max-num-seqs` +- disabling async scheduler \ No newline at end of file From 4444dcdfeadc1dd221bc6b6f65035970eea419ac Mon Sep 17 00:00:00 2001 From: Agata Dobrzyniewicz <160237065+adobrzyn@users.noreply.github.com> Date: Fri, 21 Nov 2025 07:35:41 +0100 Subject: [PATCH 2/4] Update docs/user_guide/troubleshooting.md Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> Signed-off-by: Agata Dobrzyniewicz <160237065+adobrzyn@users.noreply.github.com> --- docs/user_guide/troubleshooting.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/user_guide/troubleshooting.md b/docs/user_guide/troubleshooting.md index 3e907f1308..e9318de244 100644 --- a/docs/user_guide/troubleshooting.md +++ b/docs/user_guide/troubleshooting.md @@ -36,6 +36,6 @@ So for this specific scenario correct value for --max_concurrency is 10 #### Solution: 0.11.0 version is missing crucial fixes from upstream. To avoid it try: -- lowering max conccurency +- lowering max concurrency - decresing `max-num-seqs` - disabling async scheduler \ No newline at end of file From 48efe5888cf7b0b45b9d2ac4716815c23a4ed5b3 Mon Sep 17 00:00:00 2001 From: Agata Dobrzyniewicz <160237065+adobrzyn@users.noreply.github.com> Date: Fri, 21 Nov 2025 07:35:48 +0100 Subject: [PATCH 3/4] Update docs/user_guide/troubleshooting.md Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com> Signed-off-by: Agata Dobrzyniewicz <160237065+adobrzyn@users.noreply.github.com> --- docs/user_guide/troubleshooting.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/user_guide/troubleshooting.md b/docs/user_guide/troubleshooting.md index e9318de244..b5ee3da880 100644 --- a/docs/user_guide/troubleshooting.md +++ b/docs/user_guide/troubleshooting.md @@ -37,5 +37,5 @@ So for this specific scenario correct value for --max_concurrency is 10 #### Solution: 0.11.0 version is missing crucial fixes from upstream. To avoid it try: - lowering max concurrency -- decresing `max-num-seqs` +- decreasing `max-num-seqs` - disabling async scheduler \ No newline at end of file From f84cfd309d40b03bdac1419e36db30dbfdcfc047 Mon Sep 17 00:00:00 2001 From: Agata Dobrzyniewicz Date: Fri, 21 Nov 2025 09:37:44 +0200 Subject: [PATCH 4/4] Precommit Signed-off-by: Agata Dobrzyniewicz --- docs/user_guide/troubleshooting.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/user_guide/troubleshooting.md b/docs/user_guide/troubleshooting.md index b5ee3da880..f7d61a69cd 100644 --- a/docs/user_guide/troubleshooting.md +++ b/docs/user_guide/troubleshooting.md @@ -38,4 +38,4 @@ So for this specific scenario correct value for --max_concurrency is 10 0.11.0 version is missing crucial fixes from upstream. To avoid it try: - lowering max concurrency - decreasing `max-num-seqs` -- disabling async scheduler \ No newline at end of file +- disabling async scheduler