search

georgemarselis / openlava-macosx

Automatically exported from code.google.com/p/openlava-macosx
GNU General Public License v2.0
1 stars 0 forks source link

offer pre-fab cluster designs #115

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
types of cluster:

1 management node/login node/master batch scheduler ( 
all-cluttered-up/bioinformatician mode)

1 management node/login node, 1 master batch scheduler

1 master batch scheduler node, 1 login node, 1 management node (must figure out 
the type of auth relationship those three will have)

1 master batch scheduler node, 1 slave node (failover mode), 1 login, node, 1 
management node  (must figure out the type of auth relationship those three 
will have | metrics here will prove a lot of help figuring out how man jobs 
mbatchd can handle without giving up the spirit or overloading the scheduler 
node | heartbeat? how? )

1 master batch scheduler node, 1 slave node (loadbalancing mode), 1 login, 
node, 1 management node  (must figure out the type of auth relationship those 
three will have | metrics here will prove a lot of help figuring out how man 
jobs mbatchd can handle without giving up the spirit or overloading the 
scheduler node | heartbeat? how? )

1 master batch scheduler node, 2 slave node (failover and loadbalancing mode - 
does this make sense?), 1 login, node, 1 management node  (must figure out the 
type of auth relationship those three will have | metrics here will prove a lot 
of help figuring out how man jobs mbatchd can handle without giving up the 
spirit or overloading the scheduler node | heartbeat? how? )

write about it
   hook at the end: it is ready for you
   puppet? chef? cobbler/foreman template?

Might need to have a sponsor when we get to this

Original issue reported on code.google.com by geo...@marsel.is on 12 May 2013 at 10:28

GoogleCodeExporter commented 9 years ago 
1 master batch scheduler node, 1/2 slave node(s) (failover and loadbalancing 
mode - does this make sense?), 1 login, node, 1 management node  (must figure 
out the type of auth relationship those three will have | metrics here will 
prove a lot of help figuring out how man jobs mbatchd can handle without giving 
up the spirit or overloading the scheduler node | heartbeat? how? | is this a 
viable configuration? when would anybody need a special submit-only host? how 
would it happen? )

Original comment by geo...@marsel.is on 13 May 2013 at 10:59