My test flight application crashed while doing a disk writing operation. I saw the following error in crash log
1073.76 MB of file backed memory dirtied over 705 seconds (1522.14 KB per second average), exceeding limit of 12.43 KB per second over 86400 seconds\
After a little bit of research I found from Apple documentation that System will throw an exception if disk writes from the app exceeds a certain limit in 24 hours window
[https://developer.apple.com/documentation/xcode/reducing-disk-writes)
But the threshold is not documented. But looking at the crash log we can reverse calculate this threshold as 12.43KB * 86400 ~ 1 GB
Does this means that iOS application won't be able to write data more than this limit (1 GB) per day?
Hi,
So, let me clarify a few different points across this thread:
The report usually notes that, although it can be rather subtle. Can someone post an example of this report?
These days we have a few different file formats for storing sample traces (that's the format of the logs you're looking at) but in the original "basic" text format, most developer logs have this field near the top:
`Action taken: none
That means exactly what it sounds like- the system capture the log and then did nothing else. Note that when I say "most developer logs", I can't actually remember ANY developer log that was actually a termination.
But the threshold is not documented. But looking at the crash log we can reverse calculate this threshold as 12.43KB * 86400 ~ 1 GB
The logs are intended to be "self documenting", so the specific limit should be listed in the log. You do see cases like this:
- iPad14,3 reached 4294.97 writes
Sometimes that's because the limit changed, sometimes it's because the kernel collected the log "again" at some later point. Either way, you should be able to tell the difference using the limit referenced in the log.
Does this means that iOS application won't be able to write data more than this limit (1 GB) per day?
No, in this case it means that an iOS app won't be able to write more than ~1 GB/day... without the system generating a log about it. In more practical terms, the actual threshold is fairly arbitrary- the point of this metric is to help detect cases where unnecessary write are being generated without generating massive amounts of irrelevant noise. In that context, 1 GB/day is a pretty reasonable compromise point, as very few apps "intentionally" write that much data, particularly not "user data" that MUST be written.
Of course, there are absolutely going to be outliers that can easily exceed that limit for perfectly sensible reasons. Your app is a clear example of that.
If you say it does not terminate the app that must be the case, and if it's so, then this is not an issue in cases where the app should be doing that disk intensive tasks.
Yep. In your case, all I would really do is take a quick look at the logs to confirm they show what you're "expecting". In your case, what I'd expect to find is that the writes are all coming from specific "bulk I/O" paths and that's what I found from the spot checking I did. Assuming everything looks normal, then you can just ignore them.
-Kevin Elliott
DTS Engineer, CoreOS/Hardware
