lucgone
Member
OP
Le Baguette
- Joined:
- Jul 2024
- Posts:
- 296
- From:
- Lyon, FR
Hello, since 2 days I have make a strange problem with my VPS at Contabo. I have a systemd service for sync files, it is working very well, but after exactly 48 hours it fails. I restart, it is good again, 48 hours again, fail.
I have check:
- logs: nothing special
- memory: ok
- disk: ok
- it is possible to do that with cron? No, same problem
I am not sure what to check now. It is possible you help me? I have search since 2 days and I find nothing.
Thank you for any idea!
Vive la résistance... électrique
kate3
Member
- Joined:
- Jul 2024
- Posts:
- 208
- From:
- Utrecht, NL
After exactly 48 hours it fails
This precision suggests a virtualization tax issue, not application. OpenVZ would show different symptoms—usually OOM kills or cgroup freezes. Your 48-hour cycle points to kernel-level resource.
Check if your provider uses KSM or memory ballooning. The "exactly 48h" pattern matches a known kernel bug in 5.x where a timer overflows at that interval under certain KVM clock configurations. The virtualization tax here is clocksource stability, not performance.
What kernel?
uname -r
virsh list --all | wc -l: 47
PascalOvid
Member
- Joined:
- Jan 2025
- Posts:
- 28
- From:
- Bucharest, Romania
What kernel you run? This smell like timer bug, the server will crash at fixed interval, why 48h? Overflow in jiffies!
Check
cat /sys/devices/system/clocksource/clocksource0/current_clocksource
If tsc, change to hpet or kvm-clock, I have fix 3 VPS like this,
256 cores for a blog. why not?
