Well it’s a free world, you can choose whatever is suitable for you. I don’t consider a mb every rescan interval a tragedy. As I said, I don’t think its leanear. .
So you’re saying that if you scan every 10 seconds, in less than 10 minutes, assuming no changes to the data, syncthing has written to disk the equivalent amount in database updates?
AudriusButkevicius I’m shocked by your dismissive stance on a possibly serious issue. These details are important. I suspect this is a bug, perhaps a freeBSD problem? There’s no way 7 files at 69MB prompt a 1MB database/index update.
I am not dismissive. I think the fact that your 7 files got your to write 1MB to a disk made you trigger happy and you think you’ve discovered a bug. I think it’s more todo with number of devices rather than number of files.
Let’s talk when you have 10TB of files to verify that it’s linear. Until then, sadly your statistically irrelevant case of 7 files has no meaning in the scope of possibilities.
I sincerely hope it isn’t proportional, but I think it’s a reasonable case to have <100MB worth of a few files, and expect the service to function efficiently.
Well it’s likely that overhead of basic things we can to keep in sync is causing you suspicion, rather than something relative to the number of files.
Though regardless, I am sure other solutions cause around 1MB of disk IO to perform the same task.
Hard to believe. 1MB is a ridiculously large amount of indexing for 7 files. There must be another explanation.
And no, some other solutions use inotify, or make checksums on the data before updating the index.
As I said, I fell there is a static overhead, known as C, whereby C is constant.
If you care about 1MB of disk IO, you are definitely looking at the wrong solution.
I know what you mean. But think, even at 10TB, there’s no reasons to have a constant load on the disk. (Even if that load stays at 1MB/scan.)
Why, if there is a constant set of data we need to update?
Erm, because you don’t…
EDIT: sorry i meant ‘because you shouldn’t need to’
Why are you telling a Syncthing maintainer what data he updates? Especially after he’s told you multiple times that:
So I think you are tripping on your own arguments and not understanding what is what. Therefore there is no reason to continue any reasonable discussion with you as you can’t even get your facts straight and mistake network IO with disk IO while looking at some random command you found on stackoverflow.
The 1mb/scan has nothing to do with the network. I have paused the other devices so the FreeBSD server is running alone. Still a 1MB write to disc per scan when no files have changed.
I’ve been referring only to disk updates.
Look at your first post, you seem to have no clue what you are referring to.
The second sentence does not necessarily follow from the first. I think you make too many assumptions with too little knowledge.
The title is misleading. I thought it may have been network IO, but have learnt otherwise as this discussion has evolved. Is there a way to edit the title?
How come you’re so afraid of a possible, very niche, OS-specific bug? Isn’t it possible? You’re so quick to dismiss it. I’m surprised that you are labelled as a developer, yet so unwilling to accept the possibility of something gone very slightly wrong.
He’s one of the two people who write the vast majority of Syncthing. Be nice, please