Hello,
If I host files on a CIFS share and the connection is unstable, how does LibreOffice handle incomplete writes? What is the real risk of data corruption?
You tagged “base”, so this is a question on record-writes in databases?
I think, it doesn’t. It relies on underlying layers (OS) to handle that.
If your connection is unstable, I would assume data corruption ~inevitable.
I would not waste to much time on calculations, as I assume most figures are unknown to you.
- How “unstable” is the connection in percent of time
- How likely is there a open connection by LO? (Do you close files, or have 20 open at the same time…)
- How likely are documents changed? (A read-only document should have no risk of corruption).
- Length of record/document, as time to write is a factor in probability to be hit.
- Error recovery with checksums possible/available.
- Have you experienced connection losses? Can you still save to another place then? Does connection recovers, so you can save again later?
What I would do is working local and use a verified sync to other places. But that was not your question.
If you really wish to get figures, test your system. Use a shell script with a bunch of repeated jobs and let it run. Compare the resulting files and you will get some figures. Even then a “3 of 20000 files are not as expected” result as example may need interpretation, how typical this two files are for you to obtain “your” risc.
PS: In my case (using a database): Most databases use internal systems to check record writes and can revert, if not complete or will not accept an incomplete update (connection loss during write). But this is on writing records, not complete zip-files.
Sorry, this is for writer and calc.
Ok, I changed your tags accordingly.
.
But at the core the answer stays the same: Only you have the possibility to calculate your “real” risk. But I would prefer to put work in avoiding this risk as suggested above and add a layer of automated backups.
Afaik the integrity of a file write operation is delegated to the OS. If the CIFS connection is unstable, file corruption is almost certainly guaranteed at some stage.
In a previous professional environment, I experienced something similar with OpenOffice.org due to a bug in the NFS protocol of the version of NFS that was available at the time. It lead to lots of disgruntled users and a quite stressful time for our sysadmin.