Good morning people. I'm experiencing a weird problem.
I have a specific VM that the IO simply crashes.
This VM, during the backup period, executes mysqldump and tries to
compress, after a long period of compacting the JBD2 process appears using
100% of the CPU and also of WA (Wait in the TOP command). Numerous other
VMs perform the same process but only this one remains in this state. When
JBD2 appears, it is necessary to perform a forced reboot, because it stops
practically everything.
I'm using CephFS as Storage Backend. In the VM I can get rates of at least
600 Megs, I can generate gigs and more gigs of data without any problem.
The VMs are using Virtio instead of Virtio-blk.
Hosts are interconnected on 10G SFP+ interfaces, Storages on 40G interfaces.
I believe that the problem is not the Storage, as this behavior does not
exist in other VMs.
What I find strange is that JBD2 goes up and spikes at 100%.
What could be happening? Could it be something in the Guest itself?