-
Type:
Bug
-
Resolution: Duplicate
-
Priority:
Major - P3
-
None
-
Affects Version/s: None
-
Component/s: None
-
None
-
ALL
-
-
None
-
None
-
None
-
None
-
None
-
None
-
None
MongoDB 8.2.5 enters a persistent SIGSEGV crash loop (exit code 139, signal 11) when running
on Linux kernel 6.19. The process restarts every ~31 seconds via Docker restart policy, and
each unclean crash triggers a full WiredTiger journal recovery on the next startup.
The crash loop ran undetected for ~10 days, producing 1,705+ coredumps and 2,000+ WiredTiger
journal recovery cycles. On a btrfs filesystem, this caused ~60GB of CoW extent accumulation
(ghost storage growth of ~10GB/hr).
After downgrading to kernel 6.18.7-200.fc43.x86_64, MongoDB started cleanly and ran
indefinitely with zero crashes on identical hardware, config, and data.
Environment:
- MongoDB: 8.2.5 (tcmalloc-google allocator)
- Docker: running via docker compose
- OS: Fedora 43
- Kernel (broken): 6.19.x
- Kernel (stable): 6.18.7-200.fc43.x86_64
- Filesystem: btrfs
- Hardware: bare metal homelab server
Observations:
- `docker events` showed `die exit=139` every ~31 seconds
- `coredumpctl list` captured 1,705+ MongoDB coredumps
- THP set to `always` had no effect — crash loop continued unchanged
- Zero crashes on kernel 6.18.7 with identical MongoDB config and data volume
- MongoDB startup log on 6.18.7 showed clean starts with no journal recovery messages
- is duplicated by
-
SERVER-121912 Upgrade tcmalloc to include rseq bug fix
-
- Open
-
- is related to
-
SERVER-121912 Upgrade tcmalloc to include rseq bug fix
-
- Open
-