[kba]

This was the reason for some ephemeral-storage failures I had in k8s: For very large images, with the (default) +ping, the whole file needs to be read and for large files the cache magick uses won't fit into memory and will be dumped to $TMPDIR which I need to avoid.

Draft because I don't yet understand all the ramifications. For now I keep +ping but set MAGICK_DISK_LIMIT=0 to avoid disk writes from magick altogether and observe any problematic behavior.

[kba]

Lots of

15:20:00.317 ERROR ocrd.exif - [PHYS_0002] identify exited with non-zero 1: identify-im6.q16: cache resources exhausted

now but no failures so far.

[bertsky]

Please help me understand this:

Are you saying that the /tmp dumping behaviour (which you don't want) is controlled by MAGICK_DISK_LIMIT, and that you can already avoid the failures b setting that to zero?

And that using -ping (but no MAGICK_DISK_LIMIT) achieves the same effect?

I cannot seem to find documentation on either these variables or how they relate to the /etc/ImageMagick/policy.xml settings...