Increased memory consumption of Sentinel 1 processing since upgrade to v8.0.5

Hi all,

I’m using SNAP to calibrate Sentinel-1 level 1 data to Sigma0 and convert it to dB. This is running in a containerized setup in kubernetes.
Last week, I updated the SNAP version from v8.0.3 to v8.0.5. This version change caused the processing of some granules to crash due to memory issues.
Before the update, 1 cpu with 8 Gb of memory used to be sufficient. For now, I’ve increased the resources to 1 cpu with 12.5 Gb of memory, to ensure that the granules finish processing.

I was wondering if something changed in the latest SNAP minor version increase regarding S1 processing? I looked at the JIRA release page, but I could not find an information regarding this.

Thanks in advance.

Kind regards,

1 Like


It can be that the fix for SNAP-1382 increases the memory consumption in some scenarios. The memory cache was to often cleared. and therefor empty tiles were written.

Thanks @marpet for the swift reply.

Do you expect the memory usage to be reduced in the coming updates, or will this stay as is?

I expect it will stay like this in the near future (let say a year). In the longer term we always try to reduce memory consumption. But I can’t give you better schedule.

Thanks, knowing this helps with our resource allocation.