The system is experiencing a problem centered around OFD lock contention on a single file: /tmp/krb5cc_<UID>
Issue
- The system running RHAAP workloads is experiencing an elevated CPU %sys usage.
- Perf data shows the primary contributors to the elevated CPU usage are uwsgi and awx-manage processes, which are consuming a significant number of CPU cycles due to repeated spinlock busy-waiting.
Samples: 14M of event 'cpu-clock:pppH', Event count (approx.
Overhead Command Shared Object
+ 58.26% uwsgi [kernel.kallsyms] ◆
- 40.72% native_queued_spin_lock_slowpath ▒
- 40.72% _raw_spin_lock ▒
- 40.72% posix_lock_inode ▒
- 40.72% fcntl_setlk ▒
do_fcntl ▒
__x64_sys_fcntl ▒
do_syscall_64 ▒
entry_SYSCALL_64_after_hwframe ▒
0x7f658ca4ea3c ▒
+ 3.08% _raw_spin_lock ▒
+ 2.99% posix_lock_inode ▒
+ 1.76% do_syscall_64 ▒
+ 1.47% posix_locks_conflict ▒
+ 0.83% avc_has_perm ▒
0.79% copy_user_generic_unrolled ▒
+ 0.71% kmem_cache_free ▒
+ 0.53% kmem_cache_alloc ▒
+ 0.51% syscall_slow_exit_work ▒
+ 15.27% awx-manage [kernel.kallsyms] ▒
- 10.30% native_queued_spin_lock_slowpath ▒
_raw_spin_lock ▒
- posix_lock_inode ▒
- 10.30% fcntl_setlk ▒
do_fcntl ▒
__x64_sys_fcntl ▒
do_syscall_64 ▒
+ entry_SYSCALL_64_after_hwframe ▒
+ 0.82% _raw_spin_lock ▒
+ 0.77% posix_lock_inode ▒
+ 0.51% do_syscall_64 ▒
...
Environment
- Red Hat Enterprise Linux 8.10.z
- Red Hat Ansible Automation Platform 2.4
Subscriber exclusive content
A Red Hat subscription provides unlimited access to our knowledgebase, tools, and much more.