I changed my OH server about 250 days ago and just recognized that into the new SSD drive, there has been 12TB write in SSD hardware level. Server is dedicated to OH, so WTF? I started to dig this out and recognized that mapdb is constantly writing transaction file. I have pretty busy system about 1000 items and many of item is updated every couple of seconds. That’s why I have increased mapdb commit interval to 60 seconds (many years ago already). I was “hoping” that mapdb persistence service then writes file ones per every minute, but that’s not case. MapDB actually write every item change to transaction file and commit changes every minute to “real” db file. I don’t know how mapdb internally handles the transaction file (seems to be some kind of Write Ahead Log), but transaction file size grows all the time. I restarted mapdb bundle and after one hour, transaction file size before commit was around 2MB. File size increasing from 0 to 2MB and then after minute commit interval, file size start again from 0. So it means that mapdb writes at least 2MB every minute, which means ~3GB per day and ~1TB per year. Actual mapdb file size where latest values are stored is only ~30kB. So if that file is only written every minute, it means ~16GB per year, so there is huge difference.
Is this how mapdb persistence is planned to work or is there some kind bug?