Monitor a run
Every run page updates while the simulation runs. No tailing logs, no copying files back.
What updates
Section titled “What updates”- Thermo: each thermo column, plotted as it is written. One chart per
runcommand. - Speed: seconds per step over time, the slowest interval, and time left for the current run. Sudden slowdowns usually mean neighbor list rebuilds, I/O, or a busy machine.
- Resources: CPU, memory and threads across the process tree, sampled every few seconds, and each MPI rank’s CPU over time, so an idle rank or a lopsided decomposition shows.
- Logs: the engine’s log, standard output and standard error, followed as they grow.
The refresh rate is under Settings → Live refresh: every 1, 2 or 5 seconds.
How the data gets to you
Section titled “How the data gets to you”The copy of the input that runs flushes its thermo output on every line. On a remote machine verletrc reads the log as it grows and verlet fetches what is new through the SSH connection, a few kilobytes at a time.
When the connection drops
Section titled “When the connection drops”Nothing is lost. The run continues and verletrc keeps reading its output. On reconnect the charts fill in the gap, and the final state appears if the run ended meanwhile.
Overview
Section titled “Overview”The Overview lists what is running everywhere and a Needs attention list of failed runs.