Changes to 'fence_sanlock'

[David Teigland]

New branch 'fence_sanlock' available with the following commits:
commit 59a67bbeb30a75e874c3e9a2a6f83c6c5d997c90
Author: David Teigland 
Date:   Thu Sep 20 17:16:06 2012 -0500

fence_sanlockd: use init script

Start/stop fence_sanlockd with init script.
    It is started with -w, which means wait for options.
    Later (from cman start), fence_node -U -> fence_sanlock -o on
    calls fence_sanlockd -s to send the running daemon the
    p,i options.

Signed-off-by: David Teigland

commit f8de4dc7bb3fcfe7873dfa08428ef894aa175de8
Author: David Teigland 
Date:   Thu Sep 20 14:06:00 2012 -0500

fence_sanlock: variable alignment

- make lease size depend on disk sector size
    - clarify fence_sanlock read_leader behavior

Signed-off-by: David Teigland

commit b8c524881b9e753a8b81892c844fb9e3a4e3f9b7
Author: Fabio M. Di Nitto 
Date:   Wed Sep 19 09:12:31 2012 +0200

fence_sanlock: unify code and silence some operations

make a common function to get leader info
    redirect some errors to /dev/null

Signed-off-by: Fabio M. Di Nitto

commit 583a6f865488034cec105ab113777f6b80067485
Author: Fabio M. Di Nitto 
Date:   Wed Sep 19 08:28:22 2012 +0200

fence_sanlock: improve action_on error checking and report

Signed-off-by: Fabio M. Di Nitto

commit 7fae84b38cc2791b64e83605786166572d70cbe6
Author: David Teigland 
Date:   Tue Sep 18 13:47:00 2012 -0500

fence_sanlock: various fixes and changes

- change device arg name to path
    - add some metadata descriptions
    - in off, make request after first acquire fails
    - in off, always success after acquire works
    - in off, check if victim re-acquired cleanly
    - in off, check if already off at start
    - in status, use ^timestamp

Signed-off-by: David Teigland

commit b9e28873fbcd11ef6418758c6bf38e32524b0a76
Author: Fabio M. Di Nitto 
Date:   Tue Sep 18 10:54:40 2012 +0200

fence_sanlock: add basic metadata to integrate with ccs_config_validate

drop node="" option (unused)

TODO:
    - add proper short/long description
    - add call in spec file:

%post -n fence_sanlock
    ccs_update_schema > /dev/null 2>&1 ||:

Signed-off-by: Fabio M. Di Nitto

commit 97a190b2ffb9a70ebd1c732db46199a2c0f6ef72
Author: Fabio M. Di Nitto 
Date:   Tue Sep 18 10:37:49 2012 +0200

fence_sanlock: don't restart fence_sanlockd if it's already running

Signed-off-by: Fabio M. Di Nitto

commit 154eb0d259887605c03e295b1e187629bace58a1
Author: Fabio M. Di Nitto 
Date:   Mon Sep 17 15:10:16 2012 +0200

fence_sanlock: cleanup shell usage

add common code/sections
    improve error handling
    add a few more FIXME points as reminders

Signed-off-by: Fabio M. Di Nitto

commit 62604e2920b6c22303aff33ce0d5e21570a4e720
Author: Fabio M. Di Nitto 
Date:   Mon Sep 17 10:16:29 2012 +0200

fence_sanlockd: add missing \n

Signed-off-by: Fabio M. Di Nitto

commit d32fccf6a29acc2a0b759eb8a9736a963a9beb05
Author: Fabio M. Di Nitto 
Date:   Mon Sep 17 10:15:08 2012 +0200

fence_sanlock: add install bits

Signed-off-by: Fabio M. Di Nitto

commit 3c561fb027e9c9773e9671b755f3fa5c5e33f94c
Author: David Teigland 
Date:   Fri Sep 14 10:31:37 2012 -0500

fence_sanlock: new code

A fence agent that uses /dev/watchdog to reset hosts.
    Per-host sanlock leases on shared storage are used:
    - for hosts to detect that they are being fenced by
      someone if they haven't actually failed
    - for hosts to verify that fenced hosts have been
      reset by seeing their host_id leases expire (not
      be renewed on disk)
    - for hosts to know how long to wait after fencing
      before they can proceed, based on acquiring the
      expired lease of the fenced host

Signed-off-by: David Teigland