timeseries reader/writer workflow runs out of memory and hangs
Most recently seen with the following workflow on the yA31 dataset (dssdata.org) on 48 cores it made it through 76 files before hanging and at least the first two machines showed the pvserver processes combined using nearly 100% of the machines memory. Time steps 76 and 77 don't use that much memory alone.
- Load a file series of *.vtm files that contain only vtus.
- Merge blocks -> Threshold
- Save Data, Select pvtu, write all timesteps as file series, Appended, zlib compression
To upload designs, you'll need to enable LFS and have an admin enable hashed storage. More information