Sound like you have an I/O or caching issue, are you running on a virtual platform?
If so migrate to RAW disk, set disk cache to write-trough, disable swappiness and lower dirty cache buffers in the kernel /etc/sysctl:
Also you cannot put a lot of pressure on a vm.
Hi. No, it is hardware platform - SAN-DB-ESM. Nothing problem with io on the san level, db level, network level but once at 3-5 hours esm stopped writing to db after this check.