[[["容易理解","easyToUnderstand","thumb-up"],["確實解決了我的問題","solvedMyProblem","thumb-up"],["其他","otherUp","thumb-up"]],[["難以理解","hardToUnderstand","thumb-down"],["資訊或程式碼範例有誤","incorrectInformationOrSampleCode","thumb-down"],["缺少我需要的資訊/範例","missingTheInformationSamplesINeed","thumb-down"],["翻譯問題","translationIssue","thumb-down"],["其他","otherDown","thumb-down"]],["上次更新時間:2025-09-04 (世界標準時間)。"],[],[],null,["# Known issues\n\nThe following issues and limitations are known:\n\n- In some cases, a Parallelstore instance is created that does not\n meet the expected performance numbers. If your\n [performance testing](/parallelstore/docs/test-performance) shows that your\n instance's IOPS are insufficient, you should delete and recreate your\n Parallelstore instance.\n\n- The number of files in a directory is limited to 1.8 million. The actual\n limit may be lower and is specific to each directory, depending on the rest\n of the file system.\n\n- Some POSIX features are [not supported](/parallelstore/docs/posix-compliance).\n\n- Mmap in python and pytorch are not compatible with the `--disable-caching`\n flag in dfuse mount. If needed, remove the `--disable-caching` flag when\n mounting dfuse, but be aware of the\n [implications](/parallelstore/docs/performance#dfuse_caching) of caching\n across multiple client VMs.\n\n- [Interception](/parallelstore/docs/interception-library) is only effective\n for dynamically linked applications that use libc. IO is intercepted in\n Python, but not Go.\n\n- Once a Parallelstore instance is created, its capacity cannot be changed.\n\n- If you use both Shared VPC and VPC Service Controls, you must have the\n host project that provides the network and the service project that contains\n the Parallelstore instance inside the same perimeter for the Parallelstore\n instance to function correctly. Separating the host project and service\n project with a perimeter might cause the existing instances to become\n unavailable and might prevent the creation of new instances.\n\n- Parallelstore instances and clients can't have an IP address from the\n 172.17.0.0/16 subnet range. Parallelstore instances use the 172.17.0.0/16\n address range for internal components. As a result, the following limitations\n apply:\n\n - Clients with an IP address from the 172.17.0.0/16 range can't connect to Parallelstore instances.\n - You can't create a Parallelstore instance with a reserved IP range that overlaps with the range 172.17.0.0/16."]]