sharscho
asked on
Problem with oracle on hp-ux cluster
I have 2 hp-ux servers as a cluster and there is an oracle database running on these servers. They share Eva discs for the database. After a firmware update of the EVA, the servers came up and the db as normally was accessible through the a-node. But the db was behaving very strange and it had to be brought down. After that I rebooted the server (a-node) but when the db came up, was the db on the b-node, so a normal failover for so far. after that I brought the a-node down again and also the b-node. The a-node has been then brought up first but without th db.... I would think when the primary node comes up, the db will come up together with it but this was not the case. When I bring up the b-node, the db comes up with it and is available for use.
Question: how can I make the db come up with the a-node again and leave the b-node as a failover server in the cluster. Because now the a-node is not a failover server because when the b-node goes down, the db does not failover to the a-node. Need some commands to figure out what the problem can be. This is not a Oracle RAC it is a hp-ux cluster. I am not familiar with hp-ux. Some help is appreciated.
Question: how can I make the db come up with the a-node again and leave the b-node as a failover server in the cluster. Because now the a-node is not a failover server because when the b-node goes down, the db does not failover to the a-node. Need some commands to figure out what the problem can be. This is not a Oracle RAC it is a hp-ux cluster. I am not familiar with hp-ux. Some help is appreciated.
gheist
Some versions of products would be handy. Also what exactly you did upgrade? How?
ASKER
The update has been on the software of the eva storage system and had nothing to do with the servers. All servers that has a connection with the eva had to go down so that the update could take place on storage system.
It updates with servers working. (Maybe some IO timeout in some picky application)....
So how did you upgrade? Really?
Did you change topology? reset parameters? did FC switch reset? Gimme some dirt?
Where are your VERSIONS???
So how did you upgrade? Really?
Did you change topology? reset parameters? did FC switch reset? Gimme some dirt?
Where are your VERSIONS???
ASKER
I will answer your questions first:
The upgrade was for the OS of the controller so it had nothing to do with the FC switches and or the HP-UX servers. FC switches were not resetted.
now today I went further and looking myself. I now know that HPUX serviceguard is the cluster software on that server.
The only thing that occurred was that the primary server has been brought down while the failover server was still running. So I think that the failover server took the disks over and even when I bring down the server with the shutdown -h -y now command, it locks the disks so that when I bring up the primary server it can not mount the disks so that the disks are available on the server. Both disks have been brought down with the command shutdown -h -y now. But when the problems occurred, the database has been brought down and the server has been brought down with only the shutdown command. And all this while the failover server was up abd running.
So the solution I need is for how to bring down the failover server so that it releases the disks so that the primary node can mount the disks at boot up. What are the commands to release and attach disks to a server in a cluster node? There are no errors in the logfile only that the disks are in use when the primary node comes up. I have no serviceguard knowledge so maybe someone can help out with this.
The upgrade was for the OS of the controller so it had nothing to do with the FC switches and or the HP-UX servers. FC switches were not resetted.
now today I went further and looking myself. I now know that HPUX serviceguard is the cluster software on that server.
The only thing that occurred was that the primary server has been brought down while the failover server was still running. So I think that the failover server took the disks over and even when I bring down the server with the shutdown -h -y now command, it locks the disks so that when I bring up the primary server it can not mount the disks so that the disks are available on the server. Both disks have been brought down with the command shutdown -h -y now. But when the problems occurred, the database has been brought down and the server has been brought down with only the shutdown command. And all this while the failover server was up abd running.
So the solution I need is for how to bring down the failover server so that it releases the disks so that the primary node can mount the disks at boot up. What are the commands to release and attach disks to a server in a cluster node? There are no errors in the logfile only that the disks are in use when the primary node comes up. I have no serviceguard knowledge so maybe someone can help out with this.
Thre are some requirements regarding FC controller firmwares for each new eva version. They need some system drivers, and some system patchlevels. On those some oracle hotfixes are advised to keep it supported....
Serviceguard for oracle or serviceguard as such?
VERSION NUMBERS NOW!!!!
Serviceguard for oracle or serviceguard as such?
VERSION NUMBERS NOW!!!!
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
Do you use ServiceGuard extension for Oracle?
Do not get ridiculous - version numbers you post so far do not make sense. Please check patchlevels yourself against HP support matrices.
Do not get ridiculous - version numbers you post so far do not make sense. Please check patchlevels yourself against HP support matrices.
If you got oracle agents in service guard like me i would be able to help diagnose problem, otherwise i could offer some hand with running oracle as a normal service under service guard..... no knowing which is your problem i am unable to helo.
ASKER
Because even when I explained the problem I did not get an advice that was pointing to a solution for the problem. We had an external here in the office for another project and just telling him what happened, he could point out what it can be related to hpux. And he knows about the storage upgrade because he was involved. He said to look at 2 things and point out what had to be done to solve the problem. That is exchange of knowledge not what I experienced in this ticket.