Troubleshooting OS Issues

Answer: If you are not careful when performing an operating system update or reload, all attached LUNs can be reformatted and data on those LUNs will be removed. If the updated system includes StorNext, this could cause StorNext to no longer function.

When performing an operating system update or reload, disconnect any fibre-attached media from the system and have only local operating system-required LUNs visible. This will make sure only the required LUNs are affected.

Answer: StorNext File System cannot see all the drives you want to configure in Red Hat Linux. When Linux is installed, it defaults to only 40 disk devices when it boots up.

To address this limitation, modify the CONFIG_SD_EXTRA_DEVS setting in your Linux config file (or use xconfig under the SCSI Support tab). Then, rebuild the kernel and reboot the system.

If you require assistance rebuilding a Linux kernel, contact the Linux support resources for your installation.

Answer: On a Solaris client, you may see the following error:

In StorNext, the metadata controller and clients use an Ethernet network to exchange file system metadata. The fsmpm is a portmapper daemon residing on each StorNext File System client and server computer. Its purpose is to register an RPC identifier to the system's portmap daemon. The fsmpm publishes a well-known port where the file system (fsm) daemons register their file system name and port access number. All clients then talk to their local fsmpm to discover access information for their associated service.

Because of the importance of maintaining this connection, a heartbeat is performed over the metadata network, so if this connection is lost, a message is sent indicating a network communication problem to the fsnameservers (xxx.xxx.xxx.xxx).

Portmapper messages are logged in the nssdbg.out log file located in /usr/cvfs/debug.

System administrators should monitor the log files to make sure that connectivity is maintained.

Answer: StorNext Storage Manager fails to write to a tape drive and marks the medium as 'suspect'.

Note: This is applicable only to Red Hat RHEL 5 and SUSE SLES 10 operating systems and StorNext 3.1.x (not to 3.5.0).

When a medium is marked as 'suspect,' check if the below messages are reported in the TSM log files:

If you receive this error message, the default settings of the SCSI generic driver of RHEL 5 and SLES 10 must be adjusted. For more information about the default settings, visit Linux.org.

The following table describes the parameters in question:

Parameter	Description
allow_dio	0 indicates direct I/O is disabled, 1 indicates enabled. Quantum recommends setting this parameter to 1.
def_reserved_size	This is the default buffer size reserved for each file descriptor. Values between 0 and 1048576 are supported. Quantum recommends setting this parameter to 524288 (= 512kB).

To verify, run these commands:

If the above commands return a value of 0, this means direct I/O is disabled. Run these commands:

If the above commands return a value less than 524288, this means the buffer size is not appropriate for LTO-4 tape drives.

Verify if the TSM startup file /usr/adic/TSM/bin/TSM_control defines any of the above parameters.

Substitute the settings as seen below or add them to the startup script after the shell declaration ( #!/bin/sh) and the initial comments.

If the issue persists after making the above changes, contact Quantum Technical Support.