Clusterware resources dependencies
- From: Stefano Cislaghi <s.cislaghi@xxxxxxxxx>
- To: Oracle L <oracle-l@xxxxxxxxxxxxx>
- Date: Wed, 29 Dec 2010 08:35:17 +0100
Hi all,
I'm trying to understand what is for Oracle an 'expected behaviour'
related to this problem. I'm using clusterware 11.2.0.2 under linux
64bit.
I've configured 5 resources (r1,r2,r3,r4,r5) and dependencies. r2
depend on r1, r3 depend on r2, r4 depend on r3, r5 depend on r4.
So I suppose that during relocate of r5, resources should be taken
down in order from r5 to r1 and taken up from r1 to r5.
This does not work. CRS start with a random resource and fail, than
try another one and fail, and so on usually for 6 or 7 times and only
then is able to find the right order.
For me this is quite strange. I've contacted oracle and after 2 weeks
they said me: "this is an expected behaviour". Ok, so they want to say
the clusterware really is not able to obey dependencies in right way?!
Resources, for a test, were created in this way:
crsctl add res r1 -type cl.head.type -attr "CHECK_INTERVAL=3"
crsctl add res r2 -type cl.head.type -attr
"START_DEPENDENCIES='hard(r1)
pullup(r1)',STOP_DEPENDENCIES='hard(r1)',CHECK_INTERVAL=3"
crsctl add res r3 -type cl.head.type -attr
"START_DEPENDENCIES='hard(r2)
pullup(r2)',STOP_DEPENDENCIES='hard(r2)',CHECK_INTERVAL=3"
crsctl add res r4 -type cl.head.type -attr
"START_DEPENDENCIES='hard(r3)
pullup(r3)',STOP_DEPENDENCIES='hard(r3)',CHECK_INTERVAL=3"
crsctl add res r5 -type cl.head.type -attr
"START_DEPENDENCIES='hard(r4)
pullup(r4)',STOP_DEPENDENCIES='hard(r4)',CHECK_INTERVAL=3"
Anyone has already had a similar experience?
Thanks
Stefano
--
http://www.stefanocislaghi.eu
--
//www.freelists.org/webpage/oracle-l
Other related posts:
