ceph/TODO.org
2011-06-09 17:17:47 -07:00

1.8 KiB
Raw Blame History

TODO results gathering

TODO archive logs even if osd exited non-zero

TODO osd etc killing while test is running

TODO automatic host locking

TODO add task for ceph qa workload running

parametrized task or move every shell script into a separate task? do the workloads live in ceph.git or teuthology.git? what about workloads written in C compile on node? binary tarball has .h etc

TODO add task for kclient mount/umount task

  • kclient: [client.0]

TODO add task for rbd image creation

  • rbd.create_image: [client.0]
  • rbd.create_image: client.0: image_size: 8192

TODO add task for rbd modprobe and dev create on client

  • rbd.modprobe: [client.0]
  • rbd.dev_create: [client.0]

TODO add task for mounting fs on top of rbd

  • rbd.mkfs: [client.0]
  • rbd.mount: [client.0]

TODO add full service rbd task that wraps the lower-level tasks

  • rbd:
  • rbd: [client.0]
  • rbd: client.0: image_size: 8192

TODO make cfuse task default to all clients

  • cfuse: [client.0]
  • cfuse:

TODO make kclient task default to all clients

  • kclient: [client.0]
  • kclient:

TODO more than one autotest task call

  • autotest: client.0: [dbench, bonnie]
  • autotest: client.0: [dbench, bonnie]

TODO force cleanup on errors

for now, you can run something like this: dsh -m ubuntu@sepia70.ceph.dreamhost.com,ubuntu@sepia71.ceph.dreamhost.com,ubuntu@sepia72.ceph.dreamhost.com 'killall quiet /tmp/cephtest/binary/usr/local/bin/cmon /tmp/cephtest/binary/usr/local/bin/cosd /tmp/cephtest/binary/usr/local/bin/cmds ; fusermount -u /tmp/cephtest/mnt.*; rm -rf /tmp/cephtest'

TODO MAYBE more defaults for autotest task, so you don't need to list all clients?

  • autotest: default: dbench client.42: [dbench, bonnie]
  • autotest: dbench
  • autotest: [dbench, bonnie]