Why does qemu-kvm + ocfs2 servers hang?

Question

KVM

The third day, problems with servers periodically pop up, in the amount of 3 pieces that work on qemu-KVM + ocfs2. The ocfs2 partition is 2 TB. When recording, copying to the ocfs2 partition (It contains qcow2 virtual machine disks), freezing of servers for a few seconds was noticed. Freezing does not always occur, it is not clear for what reasons, no errors are written in the logs. But an indistinct amount of extents was noticed when using the o2info utility and filefrag:

[[email protected] ~]# o2info --freefrag 1024 /dev/dm-2
Blocksize: 4096 bytes
Clustersize: 4096 bytes
Total clusters: 524287992
Free clusters: 205368502 (39.2%)

Min. free extent: 4 KB 
Max. free extent: 129020 KB
Avg. free extent: 30904 KB

Chunksize: 1048576 bytes (256 clusters)
Total chunks: 2048000
Free chunks: 780355 (38.1%)

HISTOGRAM OF FREE EXTENT SIZES:
Extent Size Range :  Free extents  Free Clusters  Percent
    4K...    8K-  :            32            32    0.00%
    8K...   16K-  :            16            41    0.00%
   16K...   32K-  :            30           153    0.00%
   32K...   64K-  :            85           997    0.00%
   64K...  128K-  :         12084        193648    0.09%
  128K...  256K-  :          1662         58439    0.03%
  256K...  512K-  :           697         53848    0.03%
  512K... 1024K-  :           829        143810    0.07%
    1M...    2M-  :          1411        522308    0.25%
    2M...    4M-  :          1735       1129782    0.55%
    4M...    8M-  :           535        774476    0.38%
    8M...   16M-  :           636       1922638    0.94%
   16M...   32M-  :         22351     114782570   55.89%
   32M...   64M-  :           621       7352804    3.58%
   64M...  128M-  :          8497     268834190  130.90%

[[email protected] ~]# filefrag /data2/kvm/nsglog/*
/data2/kvm/nsglog/nsglog-data-back.qcow2: 882 extents found
/data2/kvm/nsglog/nsglog-data-snap.qcow2: 8214 extents found
/data2/kvm/nsglog/nsglog-root-back.qcow2: 1426 extents found
/data2/kvm/nsglog/nsglog-root-snap.qcow2: 2523 extents found
/data2/kvm/nsglog/nsglog-swap.qcow2: 255 extents found

Questions: What do percentages mean in 64M extents? Why are there >100? Who faced the problem, help, or even there is no way to unmount ocfs2.