Quantcast
Viewing all articles
Browse latest Browse all 5224

C/C++ • Re: Use of sync(2) / fsync(2) before a hard before hard shutdowns?

I have my doubts about 5 minutes of writes outstanding, unless someone has tuned it that way; or possibly a case like the file still being open and overwritten (where the metadata doesn't change until closed). ext4(5) says a default of 5 seconds for the "commit=nrsec" option, although I'm not 100% certain if that tells the entire story. Writing a new file, closing it, then renaming over the old file should, I believe, trigger a metadata write to the journal within the commit time (i.e. 5s by default), which should be after the data is written in the default data=ordered mode. That's how I interpret the documentation, anyway; I'm not an absolute expert on the subtle inner workings of ext4.

Statistics: Posted by Murph9000 — Fri Sep 06, 2024 12:07 am



Viewing all articles
Browse latest Browse all 5224

Trending Articles