test_30d()

{ cp $(which dd) $DIR || error "failed to copy dd to $DIR/dd"   for i in \{1..10}

; do
$DIR/dd bs=1M count=128 if=/dev/zero of=$DIR/$tfile &
local PID=$!
sleep 1
$LCTL set_param ldlm.namespaces.MDT.lru_size=clear
wait $PID || error "executing dd from Lustre failed"
rm -f $DIR/$tfile
done
 
rm -f $DIR/dd
}
 
So that's 128 blocks of 1M size for each iteration.  The test logs shows out of space error in the 5th iteration:
 
ldlm.namespaces.lustre-MDT0000-lwp-MDT0000.lru_size=clear
ldlm.namespaces.lustre-MDT0000-mdc-ffff88847bdc0800.lru_size=clear
ldlm.namespaces.lustre-MDT0001-mdc-ffff88847bdc0800.lru_size=clear
ldlm.namespaces.lustre-MDT0001-osp-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0000-osc-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0001-osc-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0002-osc-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0003-osc-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0004-osc-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0005-osc-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0006-osc-MDT0000.lru_size=clear
ldlm.namespaces.lustre-OST0007-osc-MDT0000.lru_size=clear
ldlm.namespaces.mdt-lustre-MDT0000_UUID.lru_size=clear
/mnt/lustre/dd: error writing '/mnt/lustre/f30d.sanity': No space left on device
67+0 records in
66+0 records out
69206016 bytes (69 MB, 66 MiB) copied, 1.38986 s, 49.8 MB/s
 sanity test_30d: @@@@@@ FAIL: executing dd from Lustre failed
  Trace dump:
  = /usr/lib/lustre/tests/test-framework.sh:6273:error()
  = /usr/lib/lustre/tests/sanity.sh:3144:test_30d()
  = /usr/lib/lustre/tests/test-framework.sh:6576:run_one()
  = /usr/lib/lustre/tests/test-framework.sh:6623:run_one_logged()
  = /usr/lib/lustre/tests/test-framework.sh:6450:run_test()
  = /usr/lib/lustre/tests/sanity.sh:3150:main()
 
This is the size of the file system AFTER the test suite:
 
df /mnt/lustre
Filesystem     1K-blocks    Used Available Use% Mounted on
/dev/sdb1       32711388 6270568  25058032  21% /

Depending on the environment in question, timing for this test may vary, which may cause us to hit OOS due to only provisioning a 400M device.  Bumping this simply to 1G provides sufficient room that this rare bug no longer recurs.

Patch to be sent shortly.