When I resume, the result from the input that worked with 10 GB is successfully retrieved from cache, but the result for a large input file that had 2 failures before success, with memory = { 10.GB * 3 } = 30.GB, is not retrieved from cache.
Increasing the initial memory so that the process works first time fixed the issue. However I don’t particularly want to allocate 30 GB for all inputs, and I want to resume from the success.
Thanks for replying so fast. No, I am not using task.memory in the scripts, but my first attempt at a stripped-down example did not reproduce the problem. I will reply again when I have managed to reproduce with something less than the full suite
Thanks, @JocelynSP. In the meantime, I’d recommend having a read at this blogpost, that shows some cases in which the process was [maybe unexpectedly, for the user] re-run. More here, and here.
I can’t make a reproducible example. It seems to be an intermittent problem, and my attempts to reproduce with a small example are not succeeding in failing.
Please close this, and if it recurs I may raise it again