이 페이지에서는 Deep Learning VM Image 인스턴스를 생성할 때 발생할 수 있는 문제와 문제 해결 방법을 설명합니다.
할당량 초과
증상: - Quota 'NVIDIA_P4_GPUS' exceeded. Limit: 0.0 in region
us-east1.
문제: 할당량이 충분하지 않습니다.
해결방법: GPU를 사용하여 인스턴스를 만들려면 먼저 GPU 할당량이 있어야 합니다.
할당량 페이지에서 프로젝트에 사용 가능한 GPU가 충분히 있는지 확인하세요. GPU가 할당량 페이지에 나와 있지 않거나 추가 GPU 할당량이 필요한 경우 할당량 상향 조정을 요청하세요. 프로젝트에 기존 결제 내역이 있는 경우에는 요청을 제출하면 할당량이 자동으로 제공됩니다. 무료 체험판 계정에는 기본적으로 GPU 할당량이 없습니다.
선점형 GPU와 일반 GPU는 따로 할당량 요청을 해야 한다는 점에 유의하세요. 선점형 GPU 할당량은 일반 GPU에 사용할 수 없습니다. 또한 리전별로 할당량이 주어지므로 할당량이 있는 리전에서 인스턴스를 만들어야 합니다.
리소스를 찾지 못함
증상: - The resource 'projects/deeplearning-platform/zones/europe-west4-c/acceleratorTypes/nvidia-tesla-p4'
was not found
문제: GPU를 사용할 수 없는 리전에서 GPU가 1개 이상 포함된 인스턴스를 만들려고 합니다(예: europe-west4-c에서 P4 GPU가 포함된 인스턴스).
일부 Linux 클라이언트에서 localhost는 IPv6 루프백 주소(::1)로 확인됩니다. ping -c 1 localhost를 사용하여 이를 확인하세요. 이 명령어가 IPv6 주소를 반환하면 gcloud compute ssh 명령어에서 -L 8080:localhost:8080 대신 -L 8080:127.0.0.1:8080을 사용합니다.
로컬 클라이언트에서 https://localhost:8080 대신 http://localhost:8080에 연결해야 합니다.
[[["이해하기 쉬움","easyToUnderstand","thumb-up"],["문제가 해결됨","solvedMyProblem","thumb-up"],["기타","otherUp","thumb-up"]],[["이해하기 어려움","hardToUnderstand","thumb-down"],["잘못된 정보 또는 샘플 코드","incorrectInformationOrSampleCode","thumb-down"],["필요한 정보/샘플이 없음","missingTheInformationSamplesINeed","thumb-down"],["번역 문제","translationIssue","thumb-down"],["기타","otherDown","thumb-down"]],["최종 업데이트: 2025-09-04(UTC)"],[[["\u003cp\u003eThis page provides solutions to common issues encountered when creating Deep Learning VM instances, such as quota limitations and resource unavailability.\u003c/p\u003e\n"],["\u003cp\u003eInsufficient GPU quota can prevent instance creation, requiring users to check their quotas and request increases, considering the distinction between preemptible and normal GPU quotas per region.\u003c/p\u003e\n"],["\u003cp\u003eInstances with GPUs must be created in regions where those GPUs are available, as attempting to create an instance in an unsupported region will result in a resource not found error.\u003c/p\u003e\n"],["\u003cp\u003ePreemptible instances cannot be created using the Google Cloud Marketplace UI, requiring the use of the CLI and the \u003ccode\u003e--preemptible\u003c/code\u003e flag.\u003c/p\u003e\n"],["\u003cp\u003eConnectivity problems with JupyterLab via SSH port forwarding can be resolved by using the correct IP address in the SSH command and connecting to the correct HTTP address.\u003c/p\u003e\n"]]],[],null,["# Troubleshooting\n\nThis page describes problems that can come up when creating Deep Learning VM Images\ninstances, and tells you how to address the problems.\n\nQuota exceeded\n--------------\n\n**Symptom** : `- Quota 'NVIDIA_P4_GPUS' exceeded. Limit: 0.0 in region\nus-east1.`\n\n**Problem**: You do not have enough quota.\n\n**Solution** : You must have GPU quota before you can create instances with GPUs.\nCheck the [quotas page](https://console.cloud.google.com/iam-admin/quotas) to\nensure that you have enough GPUs available in your project. If GPUs are not\nlisted on the quotas page or you require additional GPU quota, [request a quota\nincrease](/compute/quotas#requesting_additional_quota). If your project has an\nestablished billing history, it will receive quota automatically after you\nsubmit the request. Free Trial accounts do not receive GPU quota by default.\n\nKeep in mind that preemptible GPUs and normal GPUs require separate quota\nrequests. You can't use preemptible GPU quota for normal GPUs. Also, quota is\nper region, so be sure that you are creating the instance in the region where\nyou have quota.\n\nResource not found\n------------------\n\n**Symptom** : `- The resource 'projects/deeplearning-platform/zones/europe-west4-c/acceleratorTypes/nvidia-tesla-p4'\nwas not found`\n\n**Problem** : You are trying to create an instance with one or more GPUs in a\nregion where GPUs are not available (for example, an instance with a P4 GPU in\n`europe-west4-c`).\n\n**Solution** : To determine which region has the required GPU, see\n[GPUs on Compute Engine](/compute/docs/gpus).\n\nPreemptible instances\n---------------------\n\n**Symptom**: I can't create preemptible instance from the UI, even though I\nhave quota.\n\n**Solution** : At this time, a preemptible instance can't be created from\nGoogle Cloud Marketplace. You must use the CLI. Be sure to add `--preemptible`\nwhen setting up your new instance.\n\nUnable to use SSH port forwarding to connect to JupyterLab\n----------------------------------------------------------\n\n**Symptom** : When [using SSH port forwarding\nto connect to JupyterLab](/deep-learning-vm/docs/jupyter),\nyou are unable to connect to the instance.\n\n**Problem**: You are trying to connect to the wrong TCP socket.\n\n**Solution**:\n\n- On some Linux clients, the localhost resolves to\n the IPv6 loopback address (`::1`).\n Check this by using `ping -c 1 localhost`. If this command returns\n the IPv6 address, use `-L 8080:127.0.0.1:8080`\n (rather than `-L 8080:localhost:8080`) in the\n `gcloud compute ssh` command.\n\n- Ensure that you connect to `http://localhost:8080`\n (instead of `https://localhost:8080`) on your local client.\n\nCommunity support\n-----------------\n\nAsk a question about Deep Learning VM on [Stack\nOverflow](https://stackoverflow.com/questions/tagged/google-dl-platform)\nor join the\n[google-dl-platform](https://groups.google.com/forum/#!forum/google-dl-platform)\nGoogle group to discuss Deep Learning VM.\n\n[Learn more about getting support from the\ncommunity](/deep-learning-vm/docs/getting-support#get_support_from_the_community)."]]
