hi all, we're still in the evaluation stage of whether to go with Ocfs2 or Red Hat's GFS cluster operating systems for Oracle 10gR2 on RHEL4. the documentation for Ocfs2 is very, very minimal (is there anything more available than just the user's guide and the faq on the http://oss.oracle.com/projects/ocfs2/ page?), so i'm a little fuzzy on the possible configurations for it. we're currently running a 2-node Oracle 9.2.0.7 RAC cluster on a tru64 OS and are wanting to move to a RHEL 2-node cluster RAC environment. so my question is - does Ocfs2 allow for a 2-node cluster, since we would like to stick with that number of nodes? if one of the nodes dies, will the other node be able to detect this and keep running stand-alone until the dead node reboots, etc.? also, does Ocfs2 require a 3rd machine to administer the nodes, or can the 2 nodes each be used to administer the entire cluster? sorry, i have VERY little experience with cluster operating systems, so i apologize in advance for the newbie language used to ask the above questions! please let me know if you need further info or if what i'm asking makes no sense! also thanks for any input you have! Deborah
I have tested 2 node configuration and the various possible failures (instance, interconnect, public interface, node) and only found 1 scenario that is not catered for. Should the network card that handles the private cluster interconnect on node 1 fail (emulate with ifdown ethx), node 2 will evict itself from the cluster and reboot. Node 1 will continue to function with a deceased card. The oracle instance on node 1 will eventually fail due to the interface card not being available. After an extensive and protracted debate with Oracle support, they confirmed that this is in fact the expected and correct behaviour, as irrational as that seems. It will, however, result in the loss of the entire cluster until node 2 reboots. Note, however, that this relates to Oracle Cluster Ready Services (CRS), not OCFS2, so would probably apply to CRS on GFS as well. This also dates back a little way (12 months or so) - I am in the process of revalidating this as it is certainly a vulnerability in a 2-node RAC cluster. If anyone can correct or update this, please do so. Sorry that this is a bit off OCFS2 topic, but it's certainly relevant. ------------------------------------------- Gavin -----Original Message----- From: ocfs2-users-bounces@oss.oracle.com [mailto:ocfs2-users-bounces@oss.oracle.com] On Behalf Of D Kennell Sent: Friday, 1 February 2008 09:53 To: ocfs2-users@oss.oracle.com Subject: [Ocfs2-users] 2-node Cluster Configuration hi all, we're still in the evaluation stage of whether to go with Ocfs2 or Red Hat's GFS cluster operating systems for Oracle 10gR2 on RHEL4. the documentation for Ocfs2 is very, very minimal (is there anything more available than just the user's guide and the faq on the http://oss.oracle.com/projects/ocfs2/ page?), so i'm a little fuzzy on the possible configurations for it. we're currently running a 2-node Oracle 9.2.0.7 RAC cluster on a tru64 OS and are wanting to move to a RHEL 2-node cluster RAC environment. so my question is - does Ocfs2 allow for a 2-node cluster, since we would like to stick with that number of nodes? if one of the nodes dies, will the other node be able to detect this and keep running stand-alone until the dead node reboots, etc.? also, does Ocfs2 require a 3rd machine to administer the nodes, or can the 2 nodes each be used to administer the entire cluster? sorry, i have VERY little experience with cluster operating systems, so i apologize in advance for the newbie language used to ask the above questions! please let me know if you need further info or if what i'm asking makes no sense! also thanks for any input you have! Deborah _______________________________________________ Ocfs2-users mailing list Ocfs2-users@oss.oracle.com http://oss.oracle.com/mailman/listinfo/ocfs2-users "Warning: The information contained in this email and any attached files is confidential to BAE Systems Australia. If you are not the intended recipient, any use, disclosure or copying of this email or any attachments is expressly prohibited. If you have received this email in error, please notify us immediately. VIRUS: Every care has been taken to ensure this email and its attachments are virus free, however, any loss or damage incurred in using this email is not the sender's responsibility. It is your responsibility to ensure virus checks are completed before installing any data sent in this email to your computer."
On Friday 01 February 2008 01:23:22 am D Kennell wrote:> the > documentation for Ocfs2 is very, very minimalThe documentation is sufficient because setting up an OCFS2 cluster is quite simple. That's a feature :-)> so my question is - does Ocfs2 allow for a 2-node cluster, since we > would like to stick with that number of nodes?Yes.> if one of the nodes > dies, will the other node be able to detect this and keep running > stand-alone until the dead node reboots, etc.?Yes, if the other node is really dead, that is--it does not emit stuff on the network and does not write to the heartbeat file on the disk.> also, does Ocfs2 > require a 3rd machine to administer the nodesNo.> , or can the 2 nodes each > be used to administer the entire cluster?Yes.
The said behavior is in ocfs2 too. As in, in a 2 node setup, if the communication link breaks down, the lower numbered node will survive. SCOTT, Gavin wrote:> I have tested 2 node configuration and the various possible failures > (instance, interconnect, public interface, node) and only found 1 > scenario that is not catered for. Should the network card that handles > the private cluster interconnect on node 1 fail (emulate with ifdown > ethx), node 2 will evict itself from the cluster and reboot. Node 1 will > continue to function with a deceased card. > > The oracle instance on node 1 will eventually fail due to the interface > card not being available. > > After an extensive and protracted debate with Oracle support, they > confirmed that this is in fact the expected and correct behaviour, as > irrational as that seems. It will, however, result in the loss of the > entire cluster until node 2 reboots. > > Note, however, that this relates to Oracle Cluster Ready Services (CRS), > not OCFS2, so would probably apply to CRS on GFS as well. This also > dates back a little way (12 months or so) - I am in the process of > revalidating this as it is certainly a vulnerability in a 2-node RAC > cluster. If anyone can correct or update this, please do so. > > Sorry that this is a bit off OCFS2 topic, but it's certainly relevant. > > > ------------------------------------------- > Gavin > > -----Original Message----- > From: ocfs2-users-bounces@oss.oracle.com > [mailto:ocfs2-users-bounces@oss.oracle.com] On Behalf Of D Kennell > Sent: Friday, 1 February 2008 09:53 > To: ocfs2-users@oss.oracle.com > Subject: [Ocfs2-users] 2-node Cluster Configuration > > hi all, > > we're still in the evaluation stage of whether to go with Ocfs2 or Red > Hat's GFS cluster operating systems for Oracle 10gR2 on RHEL4. the > documentation for Ocfs2 is very, very minimal (is there anything more > available than just the user's guide and the faq on the > http://oss.oracle.com/projects/ocfs2/ page?), so i'm a little fuzzy on > the possible configurations for it. we're currently running a 2-node > Oracle 9.2.0.7 RAC cluster on a tru64 OS and are wanting to move to a > RHEL 2-node cluster RAC environment. > > so my question is - does Ocfs2 allow for a 2-node cluster, since we > would like to stick with that number of nodes? if one of the nodes > dies, will the other node be able to detect this and keep running > stand-alone until the dead node reboots, etc.? also, does Ocfs2 > require a 3rd machine to administer the nodes, or can the 2 nodes each > be used to administer the entire cluster? > > sorry, i have VERY little experience with cluster operating systems, > so i apologize in advance for the newbie language used to ask the > above questions! please let me know if you need further info or if > what i'm asking makes no sense! also thanks for any input you have! > > Deborah > > > _______________________________________________ > Ocfs2-users mailing list > Ocfs2-users@oss.oracle.com > http://oss.oracle.com/mailman/listinfo/ocfs2-users > "Warning: > The information contained in this email and any attached files is > confidential to BAE Systems Australia. If you are not the intended > recipient, any use, disclosure or copying of this email or any > attachments is expressly prohibited. If you have received this email > in error, please notify us immediately. VIRUS: Every care has been > taken to ensure this email and its attachments are virus free, > however, any loss or damage incurred in using this email is not the > sender's responsibility. It is your responsibility to ensure virus > checks are completed before installing any data sent in this email to > your computer." > > > > _______________________________________________ > Ocfs2-users mailing list > Ocfs2-users@oss.oracle.com > http://oss.oracle.com/mailman/listinfo/ocfs2-users >
thanks so much for the responses - that's exactly what i wanted to know. and good to know that setting up an OCFS2 cluster will be relatively simple, since there is very little in Oracle so far that could be categorized as simple! Deborah