runlevel and sequence point for gfs cluster infrastructure

Bastian Blank · *nix forums addict Joined: 09 Mar 2005 Posts: 84

Hi folks

I currently try to get gfs and the cluster infrastructure into a usable
state. One of the problems are the runlevel and sequence point settings
for each step. The whole thing should start before the daemons, bus this
will make it impossible to login into the system if the cluster is not
quorate (i.e. have enough active members) at startup, as ssh and inetd
are started with the other daemons.

Possiblities:
* start before general daemons:
- cluster manager: will timeout if it can't join the cluster
- io fencing: will block if cluster is not quorate
* start after daemons
- daemons which needs access to such a fs needs to be started later.
* use a system similar to dbus:
- it is possible to handle timeouts properly and just leave the fs
unmounted if the preconditions are not met.

Bastian

--
Genius doesn't work on an assembly line basis. You can't simply say,
"Today I will be brilliant."
-- Kirk, "The Ultimate Computer", stardate 4731.3

David Schmitt · *nix forums addict Joined: 25 Feb 2005 Posts: 96

On Sunday 13 February 2005 18:47, Bastian Blank wrote:

Google