BeeGFS Tier0 Unavailable - OS User Open File Limit Exceeded During Large Data Copy
Applies to
- BeeGFS parallel file system
- Linux operating systems (all distributions)
- Workloads using mpiutilcopy or similar high-file-count operations
Issue
When running a large data copy operation (e.g., mpiutilcopy) to BeeGFS Tier0, the process fails and the filesystem becomes unavailable for writes.
The following symptoms and log outputs were observed:
Application error: “Too many open files”- The file copy halts or fails
- Users are unable to copy files until open file handles drop below a threshold
Reviewing the file open limits for the user shows:
$ ulimit -n1024$ ulimit -Sn1024$ ulimit -Hn524288The user soft open file limit (
ulimit -Sn) is set to 1024.The hard limit (
ulimit -Hn) is much higher (524288).