NFS problem causing Oracle database to hang

[Madhus]

Hi,

Network connectivity to the remote data center lost. Because of that NFS file system mounted from the server located in that data center was not accessible. Though this NFS file system is not used by Oracle database anyway databases hosted on this server stopped responding. Once we got NFS file system back database started responding. Is there anything common between Oracle databse and NFS which is causing the problem?

[Luk Vandenbussche]

There must be some action under normal circumstance on the nfs filesystem.

Can you check the process under normal circumstances

fuser -cu /nfsfilesystem

[spex]

Hello Madhus,

> Though this NFS file system is not used by
> Oracle database anyway databases hosted on
> this server stopped responding.

I doubt this statement is true. Double-check your database configuration for any reference to the NFS filesystem.

PCS

[Madhus]

Hi All,

To confim again, oracle databases are not using NFS file system at all. The file system is mend for putting installation binaries. fuser -cu /filesystem did not result anything (hung) as the file system was not accessible.

Is there any thing common between NFS and oracle? NFS uses rpc, does oracle also uses rpc for anything? Can it be that rpc hung while trying for NFS and which inturn caused oracle databases not to respond

[Martin Johnson]

I agree that you can have problems without oracle acccessing the NFS file system. Any process that accesses the NFS will hang (such as "bdf"), which can also get other processes to hang.

HTH
Marty

[Dave Olker]

Hi Madhus,

Are you positive there is nothing in the Oracle environment that references the NFS mount? I'm talking about PATH variables, environment variables, etc.? Anything that would cause the Oracle process to somehow reference the down NFS filesystem would potentially lead to a hang situation if the server is down.

I've seen plenty of cases where applications that were not using an NFS filesystem at all would hang when the server went down simply because an environment variable used by the application or by the user who launched the application, somehow referenced the down NFS mount point.

Regards,

Dave

[Bill Hassell]

Unless you are willing to do everything necessary to make the NFS server MORE reliable than your server, I would not use NFS on a production system. Oracle may not be using the NFS mountpoint but there are likely several other tools and applications that are hanging, bdf and df are hanging and of course, any new logins will hang as the login code checks all mountpoints.

[Dave Olker]

One other point here:

> Network connectivity to the remote data center lost.
> Because of that NFS file system mounted from the server
> located in that data center was not accessible.

How do you know the Oracle database going non-responsive was related to the NFS mount point in the first place? It's not like the NFS server went down and your application hung, your entire connection to the remote data center was down. Could there be another connection between the local system and the remote data center that would explain your Oracle database hang?

> Once we got NFS file system back database started responding.

I assume this means once the connection between the local and remote data centers was restored things started working again. While this would certainly explain the NFS mount point working again, this might also mean any other network connection between these sites was also restored.

Is there anything specifically tying the Oracle hang with the NFS mount point?

Regards,

Dave

[skt_skt]

other than the NFS mount you mentioned any other data is accessed across this locations. Example if the data centers act as primary and DR sites and if any DB level data replication is in progress and an immediate n/w disconnectivity could cause a DB hung too.