def dst_dir_report(jobs, dstdirs, width, prefix=''):
tab = tt.Texttable()
dir2oldphase = manager.dstdirs_to_furthest_phase(jobs)
dir2newphase = manager.dstdirs_to_youngest_phase(jobs)
headings = ['dst', 'plots', 'GBfree', 'inbnd phases', 'pri']
tab.header(headings)
tab.set_cols_dtype('t' * len(headings))
for d in sorted(dstdirs):
# TODO: This logic is replicated in archive.py's priority computation,
# maybe by moving more of the logic in to directory.py
eldest_ph = dir2oldphase.get(d, (0, 0))
phases = job.job_phases_for_dstdir(d, jobs)
dir_plots = plot_util.list_k32_plots(d)
gb_free = int(plot_util.df_b(d) / plot_util.GB)
n_plots = len(dir_plots)
priority = archive.compute_priority(eldest_ph, gb_free, n_plots)
row = [
abbr_path(d, prefix), n_plots, gb_free,
phases_str(phases, 5), priority
]
tab.add_row(row)
tab.set_max_width(width)
tab.set_deco(tt.Texttable.BORDER | tt.Texttable.HEADER)
tab.set_deco(0) # No borders
return tab.draw()
def archive(dir_cfg, all_jobs):
'''Configure one archive job. Needs to know all jobs so it can avoid IO
contention on the plotting dstdir drives. Returns either (False, <reason>)
if we should not execute an archive job or (True, <cmd>) with the archive
command if we should.'''
dstdirs = dir_cfg['dst']
arch_cfg = dir_cfg['archive']
dir2ph = manager.dstdirs_to_furthest_phase(all_jobs)
best_priority = -100000000
chosen_plot = None
for d in dstdirs:
ph = dir2ph.get(d, (0, 0))
dir_plots = plot_util.list_k32_plots(d)
gb_free = plot_util.df_b(d) / plot_util.GB
n_plots = len(dir_plots)
priority = compute_priority(ph, gb_free, n_plots)
if priority >= best_priority and dir_plots:
best_priority = priority
chosen_plot = dir_plots[0]
if not chosen_plot:
return (False, 'No plots found')
# TODO: sanity check that archive machine is available
# TODO: filter drives mounted RO
#
# Pick first archive dir with sufficient space
#
archdir_freebytes = get_archdir_freebytes(arch_cfg)
if not archdir_freebytes:
return (False, 'No free archive dirs found.')
archdir = ''
for (d, space) in sorted(archdir_freebytes.items()):
# TODO: make buffer configurable
if space > 1.2 * plot_util.get_k32_plotsize(): # Leave a little buffer
archdir = d
freespace = space
break
if not archdir:
return (False, 'No archive directories found with enough free space')
msg = 'Found %s with ~%d GB free' % (archdir, freespace / plot_util.GB)
bwlimit = arch_cfg['rsyncd_bwlimit']
throttle_arg = ('--bwlimit=%d' % bwlimit) if bwlimit else ''
cmd = ('rsync %s --remove-source-files -P %s %s' %
(throttle_arg, chosen_plot, rsync_dest(arch_cfg, archdir)))
return (True, cmd)
def test_dstdirs_to_furthest_phase(self):
all_jobs = [
self.job_w_dstdir_phase('/plots1', (1, 5)),
self.job_w_dstdir_phase('/plots2', (1, 1)),
self.job_w_dstdir_phase('/plots2', (3, 1)),
self.job_w_dstdir_phase('/plots2', (2, 1)),
self.job_w_dstdir_phase('/plots3', (4, 1))
]
self.assertEqual(
{
'/plots1': (1, 5),
'/plots2': (3, 1),
'/plots3': (4, 1)
}, manager.dstdirs_to_furthest_phase(all_jobs))
# Start running archival
elif args.cmd == 'archive':
print('...starting archive loop')
firstit = True
while True:
if not firstit:
print('Sleeping 60s until next iteration...')
time.sleep(60)
jobs = Job.get_running_jobs(dir_cfg['log'])
firstit = False
archive.archive(dir_cfg, jobs)
# Debugging: show the destination drive usage schedule
elif args.cmd == 'dsched':
dstdirs = dir_cfg['dst']
for (d, ph) in manager.dstdirs_to_furthest_phase(jobs).items():
print(' %s : %s' % (d, str(ph)))
#
# Job control commands
#
elif args.cmd in ['details', 'files', 'kill', 'suspend', 'resume']:
print(args)
selected = []
# TODO: clean up treatment of wildcard
if args.idprefix[0] == 'all':
selected = jobs
else:
# TODO: allow multiple idprefixes, not just take the first