Error check to pdsh function & add field for user to specify osd devices

[xuechendi]

1. Added error check in pdsh function Currently, CBT runs pdsh command without error check, thinking if we change the function to check stderr when invoking pdsh will be more easy for users to hands on CBT and debuging
2. Adding new field in runtest.yaml to specify osd devices cceph01: [/dev/sda1, /dev/sdb1, /dev/sdc1, /dev/sdd1, /dev/sde1, /dev/sdf1, /dev/sdg1, /dev/sdh1] cceph02: [/dev/sda1, /dev/sdb1, /dev/sdc1, /dev/sdd1, /dev/sde1, /dev/sdf1, /dev/sdg1, /dev/sdh1]

If we specify osd devices in runtest.yaml, CBT will deploy to these device rather than /dev/disk/by-partlabel/osd-device-%s-data

Signed-off-by: Chendi.Xue [email protected]

Also, I notice the runtest.yaml is out of date, so I submitted a refreshed yaml I have tested it in my env to initialize ceph cluster successfully

benchmarks: radosbench: {}

cluster: cceph01: [/dev/sda1, /dev/sdb1, /dev/sdc1, /dev/sdd1, /dev/sde1, /dev/sdf1, /dev/sdg1,/dev/sdh1] cceph02: [/dev/sda1, /dev/sdb1, /dev/sdc1, /dev/sdd1, /dev/sde1, /dev/sdf1, /dev/sdg1,/dev/sdh1] clients: [cceph01, cceph02] conf_file: ceph.conf fs: xfs head: cceph01 mkfs_opts: -f -i size=2048 -n size=64k mons: cceph01: {cceph01: 192.168.5.21} mount_opts: -o inode64,noatime,logbsize=256k osds: [cceph01, cceph02] osds_per_node: 8 tmp_dir: /tmp/cbt use_existing: true user: root