Containerization WIP

Containerfile

@@ -0,0 +1,9 @@
+FROM ubuntu:22.04
+RUN apt update && apt install -y software-properties-common
+RUN add-apt-repository ppa:elmer-csc-ubuntu/elmer-csc-ppa && apt update && apt install -y elmerfem-csc
+RUN apt update && DEBIAN_FRONTEND=noninteractive apt install -y python3-scipy python3-tqdm python3-click python3-tabulate python3-gmsh python3-matplotlib python3-subprocess-tee python3-pip python3-numpy python3-pandas
+#RUN pip install --break-system-packages pyelmer
+RUN pip install pyelmer
+COPY *.py *.yml /nice-coils/
+WORKDIR /nice-coils
+VOLUME /mesh

build_container_tarball.sh

@@ -0,0 +1,4 @@
+#!/bin/sh
+
+podman build -f Containerfile --label nice-coils-make --tag nice-coils-runner --iidfile nice-coils-runner.cid --cache-ttl 1h
+podman save --format oci-archive -o nice-coils-runner.tar.gz $(cat nice-coils-runner.cid) 

sim_runner.py

@@ -1,11 +1,14 @@
 #!/usr/bin/env python3
 
 import threading
+import datetime
 import queue
 import itertools
 import pathlib
 import tempfile
+import secrets
 import sys
+import uuid
 import sqlite3
 import time
 import math
@@ -17,9 +20,10 @@ import click
 from tabulate import tabulate
 
 
-def mesh_args(db, coil_id, mesh_type, mesh_file, outfile):
+def mesh_args(db, coil_id, mesh_type, mesh_file, outfile, **kwargs):
     mesh_type = {'split': '--mesh-split-out', 'normal': '--mesh-out', 'mutual': '--mesh-mutual-out'}[mesh_type]
-    rows = db.execute('SELECT key, value FROM results WHERE coil_id=?', (coil_id,)).fetchall()
+    rows = dict(db.execute('SELECT key, value FROM results WHERE coil_id=?', (coil_id,)).fetchall())
+    rows.update(kwargs)
     args = ['python', '-m', 'twisted_coil_gen_twolayer', mesh_type, mesh_file, '--pcb']
     for k, v in rows:
         prefix, _, k = k.partition('.')
@@ -169,7 +173,8 @@ def run(ctx, run_id, log_dir, mesh_dir):
     if run_id is None:
         run_id, = ctx.obj['db_connect']().execute('SELECT run_id FROM runs ORDER BY timestamp DESC LIMIT 1').fetchone()
     ctx.obj['run_id'] = run_id
-    ctx.obj['log_dir'] = pathlib.Path(log_dir)
+    log_dir = ctx.obj['log_dir'] = pathlib.Path(log_dir)
+    log_dir.mkdir(exist_ok=True)
     ctx.obj['mesh_dir'] = pathlib.Path(mesh_dir)
 
 
@@ -260,6 +265,73 @@ def self_inductance(ctx, num_jobs):
         threads.append(t)
     q.join()
 
+@run.command()
+@click.option('target_hosts', type=click.Path(exists=True, dir_okay=False, path_type=pathlib.Path), help='File with one SSH target host name per line')
+@click.option('job_file', type=click.Path(exists=True, dir_okay=False, path_type=pathlib.Path), help='JSON job description file')
+@click.pass_context
+def run_mutual_inductance(ctx, target_hosts, job_file):
+    db = ctx.obj['db_connect']()
+    target_hosts = [l.strip() for l in target_hosts.read_text().splitlines()]
+    log_dir.mkdir(exist_ok=True)
+    jobs = json.loads(job_file.read_text())
+
+    q = queue.Queue()
+    def queue_worker(host):
+        try:
+            while True:
+                case_name, case_id, kwargs = q.get_nowait()
+                workdir = pathlib.Path('/mesh') / f'{case_name}-{case_id}' # container path
+                mesh_file = wordir / 'mesh.msh' # container path
+                host_log = lambda log_name: ctx.obj['log_dir'] / f'{case_name}-{log_name}-{case_id}.log'
+                mesh_log = host_log('mesh') # host path
+                sim_log = host_log('sim') # host path
+                try:
+                    args = mesh_args(db, coil_id, 'mutual', mesh_file, str(workdir / 'tmp_pcb.kicad_pcb'), **kwargs)
+                    res = subprocess.run(['ssh', host, 'podman', 'run', '-v', './mesh:/mesh:rw', 'nice-coils-runner', *args], check=True, capture_output=True, text=True)
+                    mesh_log.write_text(res.stdout + res.stderr)
+
+                    try:
+                        tqdm.tqdm.write(f'Processing {mesh_file} on {host}')
+                        res = subprocess.run(['ssh', host, 'podman', 'run', '-v', './mesh:/mesh:rw', 'nice-coils-runner', 'python', '-m', 'coil_parasitics', 'mutual-inductance', '--sim-dir', workdir, mesh_file], check=True, capture_output=True, text=True)
+                        sim_log.write_text(res.stdout + res.stderr)
+
+                        try:
+                            for fn in ['ElmerGrid_stdout.log',
+                                       'ElmerGrid_stderr.log',
+                                       'ElmerSolver_stdout.log',
+                                       'ElmerSolver_stderr.log']:
+                            subprocess.run(['scp', f'{host}:mesh/{workdir.name}/{fn}', str(host_log(Path(fn).stem))], check=True, capture_output=True, text=True)
+
+                        except subprocess.CalledProcessError as e:
+                            print(f'Error copying simulation logs, rc={e.returncode}', file=sys.stderr)
+                            sim_log.write_text(e.stdout + e.stderr)
+
+                    except subprocess.CalledProcessError as e:
+                        print(f'Error running simulation, rc={e.returncode}', file=sys.stderr)
+                        sim_log.write_text(e.stdout + e.stderr)
+
+                except subprocess.CalledProcessError as e:
+                    print('Mesh generation failed with exit code {e.returncode}', file=sys.stderr)
+                    mesh_log.write_text(e.stdout + e.stderr)
+
+                tq.update(1)
+                q.task_done()
+        except queue.Empty:
+            pass
+
+    run_id = f'{datetime.datetime.now():%Y%m%d-%H%M%S}-{secrets.token_urlsafe(6)}'
+    for i, (job_name, kwargs) in jobs.items():
+        case_id = f'{run_id}-{i}'
+        q.put((job_name, case_id, kwargs))
+
+    tq = tqdm.tqdm(total=num_meshes-num_completed)
+    threads = []
+    for host in target_hosts:
+        t = threading.Thread(target=lambda: queue_worker(host), daemon=True)
+        t.start()
+        threads.append(t)
+    q.join()
+
 @run.command()
 @click.pass_context
 def self_capacitance(ctx):