Before You Install | To Install | Resources | Index

On this page:

Caveats to Read Before You Upgrade
- All-Platform
- Platform-Specific
Caveats to Read After You Upgrade

Caveats to Read Before You Upgrade

These caveats make you aware of changes or potential problems that could significantly affect your installation.

Different caveats will be relevant for different types of users; scan the list for information that is applicable for you.

This is not a complete list of all known bugs, workarounds, or changes, which are documented in the release notes for each individual product. In addition to reviewing this list, you may want to check "Caveat and Release Note Updates" link at http://support.sgi.com/6.5/caveat_updates.html for any late-breaking information.

All-Platform Caveats to Read Before You Upgrade

Confirm that the applications you rely on are supported
(This is also Step 3 of the Preparation Checklist.)

Note: An application supported on any release in the IRIX 6.5 family will run on all subsequent IRIX 6.5 family releases. Additionally, the application will be compatible with all prior releases of the IRIX 6.5 family, as long as the application does not utilize a new feature from a later 6.5.x release.

Here are some examples (not a complete list) of applications that either need to be upgraded or are not yet supported:

- DFS Domestic: DFS Domestic version 1.2.2a (1275426100) is incompatible with IRIX 6.5.12. DFS Domestic version 1.2.2c is available and is compatible with IRIX 6.5.12. DFS Domestic users must remove DFS Domestic version 1.2.2a before they upgrade and install DFS Domestic version 1.2.2c after they upgrade.

- Netscape Communicator or Navigator: By default, version 4.76 will install the new N32 version of the browser. Any existing third-party plugins for the browser that are O32 will no longer function. If you require existing O32 plugins, you can install an O32 version of the browser from netscape.sw.o32_client and netscape.o32_plugin.*. Macromedia is planning to release an N32 version of the Flash plugin in the near future. Real.com has released an N32 version of their plugin. Check the third-party web sites for updates on the status of their plugins.

- REACT/Pro: Version 3.2 is incompatible with IRIX 6.5.2 or later releases. REACT/Pro users need to upgrade to version 4.1 to run frame scheduler applications on IRIX 6.5.2 or later.

- HIPPI: Version 3.2 or later is needed to run on IRIX 6.5.x releases.To find out how to upgrade, see www.sgi.com/software (under Software Categories, click Networking, then click HIPPI 3.3.1).

- iFOR/LS: Contact the Isogon Corporation at http://www.isogon.com for the latest release of LicensePower/iFOR (formerly known as iFOR/LS) that runs on IRIX 6.5.

- Inventor: Version 2.1.6 or later is needed to run on IRIX 6.5.12. The Inventor overlay 2.1.6 is available from the /CDROM/dist directory of IRIX 6.5.12 Overlays CD (3 of 3), May 2001. This overlay requires that the base Inventor 2.1.4 image be installed already, or with the overlay during the same install session. The Inventor overlay 2.1.6 can be installed on IRIX 6.5.5 and later.

If you have Inventor 2.1.5 installed through some other mechanism, you will get a conflict when you attempt to install this release. If you get this conflict, you can resolve it by using the following set of commands. You will need your Inventor 2.1.4 installation CDROM or a network distribution for it as part of the process. From within your installation session, enter the following commands:

open <Inventor 2.1.6+ from /CDROM/dist/inventor_657+>
open <Inventor 2.1.4 from the base IRIX 6.5 CD>
install inventor_eoe
set neweroverride on

Ensure that you have not selected any other downgrades before proceeding or an incorrect installation may occur. You can check for other downgrades by using the following command:

list i D

If any subsystems other than inventor_eoe are listed, you must deselect them before you continue.

- Intranet Junction: Version 1.0.2 or later is needed to run on IRIX 6.5.x releases.

- ClearCase: See the RATIONAL Software site for information on purchasing a version of ClearCase that runs on IRIX 6.5.

- Sybase: Contact Sybase, Inc. for information on purchasing a version of Sybase that runs on IRIX 6.5.

If you have edited configuration files, check for changed versions at end of installation
Ideally, configuration file differences should be merged prior to restarting the system at the end of an installation. Before restarting the system at the end of your installation, use the Inst command admin config changed to find out which system configuration files were modified since installation. You can find out more about this command by entering help config at the Inst prompt. To avoid compatibility problems, merge configuration files if new versions were created. The "Updating Configuration Files" section of the versions(1M) man page describes configuration file updating, as does the online IRIX Admin: Software Installation and Licensing guide.

One example of a file that may need to be merged is the /etc/group file. If you have edited your /etc/group file, see the Default group file now reinstalled caveat under Caveats to Read After Upgrading.

Users who have installed ftn_eoe.sw.lib from the MIPSpro 7.3 product before installing the IRIX 6.5.x images may see a downgrade conflict with distribution subsystem ftn_eoe.sw.libeag when they install with the install maint command if they also have the original 6.5 Foundation-1 CD open when they do the installation.
If this downgrade conflict occurs, you must specify the keep ftn_eoe.sw.libeag command. The conflict occurs because ftn_eoe.sw.libeag was replaced by the 7.3 ftn_eoe.sw.lib, and the install program logic with install maint now automatically selects downgrades.

Customers who have the MIPSpro 7.3 compilers installed or who will install MIPSpro 7.3 on R4000 or R4400 platforms running IRIX 6.5.5 should install the new 7.3.1m runtime libraries that are available from Supportfolio at http://support.sgi.com.

Customers enrolled under a valid support contract who do not have web access can obtain these libraries on CD media free of charge upon request from their local support organization.

Cumulative for 6.5 through 6.5.11, and still applicable for 6.5.12:

CXFS and IRIX FailSafe coexecution configuration
As of IRIX 6.5.12f, a cluster can be configured with a maximum of 16 nodes. All the nodes in the cluster are CXFS nodes. As many as 8 nodes can also run IRIS FailSafe (coexecution).
The Fortran BLAS libraries packaged under ftn_eoe.sw.libblas (ftn_eoe.sw64.libblas on IRIX64 systems) are marked for default installation
As a result of this packaging, you may receive the following conflict during installation if you did not already install these libraries:

Overlay product ftn_eoe.sw.libblas (1276765510) cannot be installed because of missing prerequisites: base product ftn_eoe.sw.libblas (1274567300):
1a. Do not install ftn_eoe.sw.libbals (1276765510)

1b. Also install base product ftn_eoe.sw.libblas (1274567300) from an additional distribution -- insert another CD or specify another software distribution

You can resolve this conflict by either not installing libblas or by installing the base product from the IRIX 6.5 Foundation 1 CD.
System processor compatibility with XVM
XVM installed with CXFS is currently qualified on IP27 systems (Origin 200, Origin 2000, and Onyx2), IP30 systems (Octane) and the SGI Origin 3000 server series.
XVM can be used only when running CXFS
XVM is the new volume manager that is required when you use the CXFS filesystem. XVM can be used only when it is layered with the CXFS filesystem, except if you are using a system (root/swap/usr) disk. Other than for system disks, XVM is not currently supported as a standalone volume manager to replace XLV. If CXFS is not being used on a system, SGI will support only XLV as a standalone volume manager. The current required licensing for CXFS includes the use of XVM with CXFS. XVM is only available on the IRIX feature stream.
Use the XVM give and steal commands with extreme caution
XVM runs in either the local domain or in the cluster domain, and XVM disks are classified according to whether they are owned by the local domain, an active cluster domain, or a foreign domain. The XVM give and steal commands let the administrator change the ownership of a list of XVM disks. Here are some guidelines to follow when using the give and steal commands:
- If you are changing the domain ownership of XVM disks, you should use the give command instead of the steal command. The steal command is intended to be used only as a last resort when the give command cannot be used.
- In general, an XVM probe command should be run after the give command. If the XVM disk was given to a local host, the probe command is run on the new owner host. If the XVM disk was given to a new cluster, the probe command is run on any host that is a member of that cluster.
- The give and steal commands should never be used so that a multidisk volume configuration is split between domains. The user must ensure that the list of disks that will move from one domain to another is complete; that is, all volume configurations are fully contained on this set of disks. The XVM show command can be used to examine the top-down configuration for all volumes in the system, including the full set of disks that contain parts of these volumes.
The NUMA migration call migr_range_migrate() was not reporting a migration failure
The NUMA migration call migr_range_migrate() was not reporting a migration failure when the migration target was either a nonexistent node or an uninitialized policy module of an SGI Origin system. Beginning with the IRIX 6.5.9 release, this error condition is now reported back to the user application as EINVAL. For more information, see the migration(5) man page. This caveat will affect the MediaBase application and it will be fixed in a subsequent release.
The fibre channel storage area network (SAN) port names may change
There has been a change in the way the final component of a fabric target's path will be generated. Some paths that include a 16-digit port name in the final component may have that port name shortened to a single digit port number. If you use fibre channel switches to form SANs, this change will affect you. This name change will require you to update the configuration files that contain the paths to fibre channel fabric devices, including the following files:
- /etc/fstab
- /etc/failover.conf
You are not impacted by this change if the paths to your SAN targets are similar to the following:

/dev/dsk/5006094670006f7d/lun0s7/c4p1

You may be impacted by this change if the paths to your SAN targets are similar to the following:

/dev/dsk/5006094670006f7d/lun0s7/c4p5006094670006f7d

The path name to the target might be shortened as follows:

/dev/dsk/5006094670006f7d/lun0s7/c4p1
Miser unable to schedule jobs during the transition from Daylight Savings Time to Standard Time
If the requested time of a submitted Miser job is such that the job, when scheduled, will be executed during the transition time period from Daylight Saving Time back to Standard time, Miser will not schedule the job until after the transition occurs.
Installing CXFS and XVM products available in the IRIX 6.5.11 release
If you install the CXFS and XVM products that are available in the IRIX 6.5.11 release, it is highly recommended that you check the "Caveat and Release Note Updates" link at http://support.sgi.com/6.5/caveat_updates.html for information on the latest CXFS and XVM patches, CXFS and XVM feature descriptions, and any late-breaking caveats.
The miser_qinfo command only reports the first 250 scheduled jobs
This limitation is due to the Miser buffer limit and will be fixed in a subsequent release.
The miser_qinfo command rounds up the amount of reported available system memory
The amount of reported system memory available may be inaccurate by .5 gigabytes (GB) when Miser is configured with more than 1 gigabyte (GB) of memory.
Repacking stops scheduled Miser jobs if daemon is restarted
If scheduled Miser jobs are in a Miser queue that has a repack scheduling policy, after Miser has been stopped and restarted, the repacking stops for the existing jobs.
Submitting a large number of jobs to a Miser queue
If you submit a large number of jobs to a Miser queue, it may result in some of these jobs being hung and failing to terminate.
Coexecution of CXFS and IRIS FailSafe
CXFS 6.5.12f can install and run CXFS with IRIS FailSafe 2.1 or later, with applicable licensing. This is known as coexecution. However, CXFS cannot reside on the same system with earlier versions of IRIS FailSafe. (If you first install an earlier version of IRIS FailSafe, you must deinstall it before you can install CXFS, and vice versa).

XVM is currently provided at no additional charge with this release. With future releases, XVM will be repackaged, licensed, and supported as an additional software product for use with the IRIX operating system. For more information on how to obtain a license, contact your SGI service provider.
Xservers file changes
Starting with 6.5, the file /var/X11/xdm/Xservers is no longer automatically replaced when installing new versions of x_eoe.sw.eoe. Instead, the new version is installed as Xservers.N. This allows most users to preserve changes that they have made to their Xservers file without having to merge them back in with each new software upgrade.

Also, the Xsgi command line argument shmnumclients is no longer supported, and will prevent the X server from coming up if it's specified in the Xserver file. If you added the shmnumclients flag to the Xservers file, remove the shmnumclients flag before upgrading to 6.5 or later.
"Filesystem too large for device" Error Message
"Filesystem too large for device" error message
If you have repartitioned your disk drive with fx but did not create a filesystem with mkfs, you may see a message like this (under certain rare conditions) when booting the miniroot:

"Mounting file systems:
mount: /dev/dsk/dks0d1s0 on /root: Filesystem too large for device.
mount: giving up on: /root
Unable to mount partition: /dev/dsk/dks0d1s0 on /root.
This is your system disk: without it we have nothing on which to install software.
Please manually correct your configuration and try again.
Press Enter to invoke C Shell csh:"
This generally indicates that you need to run /sbin/mkfs on the named disk partition. Inside the miniroot, a convenience script (mrmkfsrc) is provided to run mkfs.

For example, for the above failure, you could press Enter and do the following (only if the disk is free of important data, because this procedure will remove all the data on the disk):

# mkfs /dev/dsk/dks0d1s0 (if this is the system disk)

# exit

(To exit csh and have the miniroot script try to remount root.)

Upgrade the flash prom after installation on diskless systems (see under O2-specific caveats and Octane-specific caveats)
During miniroot installation: innocuous Appletalk error messages
If you install from the miniroot, the following message may be displayed during the installation of Appletalk. You can safely ignore it.

"Command "( if test `grep ksd $rbase/var/adm/appletalk/services | wc -l` -eq 0; then echo "# Start K-AShare Server\\n/usr/etc/appletalk/ksd -f -o -i X/usr/bsd/hostnameX" | tr X \\140 >> $rbase/var/adm/appletalk/services; fi )" failed (return status 127, subsystem appletalk.sw.xinet_base 1429002000)"
Obsolete software may cause installation conflicts
Starting with the 6.5.3 release, the installation process detects some additional software package versions that are obsolete and incompatible with the 6.5 release family. During installation, you may see conflict messages caused by such software. When you see these conflict messages, simply choose to remove the incompatible software or replace it with newer versions (if available).
MediaMail now replaced by Netscape mail
MediaMail is no longer available from SGI. The Netscape mail program, which is bundled with Netscape Communicator, is the new default installed e-mail program for all SGI systems. SGI recommends a transition to Netscape e-mail (Netscape Messenger) and have included software with Netscape that enables conversion of MediaMail mail to Netscape e-mail format.
When you transition from MediaMail to Netscape mail:
- Be sure to check your home directory for a .forward file. If you have one that contains a line that mentions /usr/lib/Zmail/bin/reassembler, remove or rename the .forward file immediately before or after the IRIX installation, in order to prevent problems with mail delivery to your account. On multi-user systems, each user should check for this file within their own home directory.
- To learn about getting started with Netscape Mail, see About Netscape Messenger (you can also access this page by choosing Find > WebTools from the Toolchest, and double-clicking the AboutNetscapeMessenger icon).
Note: If you still require MediaMail, please contact NetManage, Inc. for product and sales information: http://www.netmanage.com. (Note that NetManage's Z-Mail for UNIX product is the same as MediaMail.) Please be aware that MediaMail is no longer officially supported on IRIX.
Appletalk will not work on diskless systems
Appletalk (Xinet Macintosh Connectivity) won't work on a diskless system because appletalk installs files into /usr/adm.
XFS filesystem is strongly recommended
To install this version of IRIX you should have an XFS filesystem. If you want to move a disk with a 6.5 (or later) filesystem to a pre-6.5 system, use the -d unwritten=0 option when you run mkfs, or install the current XFS patches on the pre-6.5 system.
During miniroot installation: innocuous disk space error message
You may may ignore the following message if you see it during a miniroot installation:
"/proc/pinfo/: no such file or directory"
The ftn_eoe.sw.libblas and ftn_eoe.sw64.libblas are no longer marked for default installation.
This caveat applies only to the IRIX 6.5 through IRIX 6.5.9 releases; it is not applicable to the IRIX 6.5.10 and later releases. If you require these subsystems, you must specifically select them during installation.
If you are upgrading from a pre-6.5 release or to a clean disk, you must open the core 6.5 CDs during the installation.
The instructions in the Installation Instructions CD booklet will prompt you to do this at the right time.

Platform-Specific Caveats to Read Before You Upgrade

If you have an Indigo R4000, Indigo2, Octane, O2, Origin 2000, or Origin 3000 please check the relevant caveats below before you begin your installation.

Indigo R4000 and Indigo2 R4000 Systems

If the system you are upgrading has a separate filesystem for / and /usr, then the / filesystem must be at least 25MB in order to do the upgrade. Systems shipped with IRIX 4.0.x were shipped with separate / filesystems of approximately 16MB, which is too small for this version of IRIX.

To check, use the command "df -k /"

If the df -k / output looks similar to

`Filesystem`	`Type`	`kbytes`	`use`	`avail`	`%use`	`Mounted on`
`/dev/root`	`efs`	`15275`	`9492`	`5783`	`62%`	`/`

the disk will need to be backed up, and repartitioned with the fx command before IRIX can be installed.

To do this, refer to the online book IRIX Admin: Software Installation and Licensing, Troubleshooting section, Using fx to Restore the Swap Partition, for how to do this with the fx command. However, before you begin this procedure, make sure you back up important data! And, instead of the last step, label/create/boot, use the command /repartition/root, then press Enter in answer to the prompt:

"fx/repartition/rootdrive: type of data partition = (xfs)"

Next, when you see

"Warning: you will need to re-install all software and restore user data from backups after changing the partition layout. Changing partitions will cause all data on the drive to be lost. Be sure you have the drive backed up if it contains any user data. Continue?"

press Enter.

The kbytes number should be at least 25600 in order to install IRIX, and it is recommended that it be at least 40MB, particularly on systems that support running 64 bit address executables (systems where uname -s returns IRIX64, on systems running IRIX 6.2 and later).

Indigo2 Systems

The EISA Token Ring driver is not supported.

Octane Systems

Upgrade the flash prom after installation on diskless systems
On diskless systems, the installation may fail if the flash prom is too old. To upgrade it on an Octane, enter the following lines in a shell:

su
/usr/sbin/flash -P /usr/cpu/firmware

You can compare the before and after results of /usr/sbin/flash -V to confirm the revision change. Reboot afterward.

O2 Systems

Upgrade the flash prom after installation on diskless systems
On diskless systems, the installation may fail if the flash prom is too old. To upgrade it on an O2, enter the following in a shell:
su
/sbin/flashinst -T -y -q /usr/cpu/firmware/ip32prom.image

You can compare the before and after results of /sbin/flashinst -v to confirm the revision change. Reboot afterward.

If you program device drivers for the O2 PCI expansion slot, you must use the routines (pciio_pio_*()) described in Chapter 22 of the 6.5 Device Driver Programmers Guide, under the section"PCI Drivers for the IP32 (O2) Platform," to perform PIO (Program IO) accesses. You must use the pciio_pio_* routines for all PIO access to the device, including the PCI config space. These routines were made available in 6.3 via patch 2801 (if your PCI device driver is already using these routines then there is nothing more you need to do).

Origin 2000 Systems

Special upgrade instructions for SGI Origin 2000 systems with 65+ processors
There are special upgrade instructions for SGI Origin 2000 systems that have 65 or more processors. To receive a copy of these instructions, please send a request to techpubs@sgi.com, specifying the kind of system you have and whether you are going to be upgrading to 6.5.12 or installing 6.5.12 onto a clean disk.
System now reboots on panic without stopping at the PROM menu
Rebooting on panic without stopping at the PROM menu is the new default behavior. You can stop the system during reboot by pressing the Esc key at the "Starting Up The System" message.
If the system panics during the reboot, the system will go into a panic/reboot loop. If this should happen, press the Esc key to break the loop.

If you do not want the system to reboot on panic, you can override the default reboot_on_panic variable with the PROM variable rebound. To do this, enter the following at the PROM prompt, accessed via option 5 at the PROM menu:

setenv rebound n
To get the MMSC version number, check the bottom line of the LCS control panel.
MMSC version numbers for this version of IRIX operating system should be 1.2.E (or higher). SGI recommends that you upgrade to version 1.2.E if your version is lower.
A second way to get the version of all the MMSCs is to connect to the alternate console port on one MMSC and enter the following:

^T track * ver
Potential FLEXlm licensing problems
The Flexible License Manager (FLEXlm) controls the use of certain software products that are asynchronous to the IRIX operating system.
A FLEXlm key is generated for each asynchronous software product, based on a unique system identifier. Because the Origin 2000 system does not have a unique system identifier, the FLEXlm host ID (lmhostid) of the first system module is used to create a FLEXlm key.

If an Origin 2000 system is reconfigured, a different module with a different FLEXlm host ID could be located in the first position, causing a previously valid FLEXlm key to fail.

If a license key problem arises after a system has been reconfigured, a new license must be generated. Contact the vendor of the software product in question and give them the host ID number (the FLEXlm host ID of the first module) of the reconfigured system to generate a new key.

SGI Origin 3000 server series

If the system you are upgrading is partitioned, the /var/sysgen/system/irix.sm file must be edited again after the installation has completed. Merge the changes that you previously made to this file back into the new version of the file to reenable system partitioning support.

Caveats to Read After You Upgrade

These caveats make you aware of changes or potential problems that could significantly affect your experience after your upgrade.

Different caveats will be relevant for different types of users; scan the list for information that is applicable for you.

This is not a complete list of all known bugs, workarounds, or changes, which are documented in the release notes for each individual product. In addition to reviewing this list, you may want to check "Caveat and Release Note Updates" at http://support.sgi.com/6.5/caveat_updates.html for any late-breaking information.

New for 6.5.12:

Netscape FastTrack Personal Web Server replaced by the SGI Web Server based on Apache
The new default web server on the IRIX 6.5.12 Applications CD is the SGI web server based on the Apache web server, version 1.3.17. Functionally, the SGI web server replaces the Netscape FastTrack web server. If you still have Netscape FastTrack enabled through the chkconfig command, the new SGI web server will not start up. To enable the new SGI web server, you must enter the following commands before you reboot your system:

# chkconfig nss_fasttrack off
# chkconfig sgi_apache on

If you changed the default Netscape FastTrack configuration, you must manually move those changes into the new SGI web server. For more information, see the SGI release notes about the SGI web server and the Apache web site (httpd://httpd.apache.org/docs/).
If you use the Web Setup product, you must turn off the Netscape FastTrack version to run the SGI version, by entering the following commands before you reboot your system:

# chkconfig webface off
# chkconfig webface_apache on

The SGI web server includes software that was developed by the Apache Software Foundation (http://www.apache.org).

This release contains no administrative GUI for the SGI web server; this will be included with a future IRIX release. For information on how to configure and administer the Apache web server, see httpd://httpd.apache.org/docs/. For a list of frequently asked questions, see httpd://httpd.apache.org/docs/misc/FAQ.html.

Cumulative for 6.5 through 6.5.11, and still applicable for 6.5.12:

Downgrading using a live (non-miniroot) installation
If you are downgrading from feature stream 6.5.x to 6.5 using a live (non-miniroot) installation, inst/swmgr will prompt you to continue installing from the miniroot. The prompt is as follows:

This software must be installed from the miniroot. The system is about
to be shut down and automatically booted to the miniroot where the
installation will continue automatically. Okay to proceed?

It is not necessary to perform this type of installation from the miniroot. To avoid this prompt, the user should set the "live_install" preference to "on" before typing go/pressing start.

If you want to downgrade to an earlier 6.5.x version of the feature stream, use the following set of commands:

set newoverride on
k *
install standard
install downgrades
install prereqs
keep incompleteoverlays
Default group file now reinstalled
With release 6.5.3, the default group file, /etc/group, is reinstalled. The original group file is saved as group.O, which appears only if you have previously edited /etc/group. This update is necessary because some optional packages, such as nss_fasttrack, require the group "nobody" to be present during the installation process.

After installation, users who have defined new groups will have to merge the old group file into the new group file. When merging, please be sure to preserve the entry for "nobody."

The Inst command admin config change and the post-installation command versions changed will both report any changed configuration files, including the /etc/group file. As mentioned in the Caveats to Read Before You Upgrade, this should ideally be done before at the end of the installation process, but before you restart the system.
New file for setting NIS server
It is no longer possible to set the NIS server in the /etc/config/ypbind.options file. If your system is configured this way, create the file
/var/yp/binding/<DOMAIN>/ypservers

where <DOMAIN> is the NIS domain name. Type the NIS server on a new line in this file.

There were several other NIS-related changes in 6.5 that affect customers who upgrade from a pre-6.5 release of the IRIX operating system. For example, the IRIX name services have been completely rewritten. Please refer to the IRIX eoe release notes (chapter 3, Changes and Additions) for details.
Remote host display now off by default
In the interest of security, the ability to display applications that are running on other systems is now disabled by default. There are a few different ways you can re-enable the remote host display. For details, see "About the Remote Display Feature" in Chapter 13 of the online Desktop User's Guide.
Printer may not be recognized if turned on after system
Even after a printer is configured, it may not function if you turn it on after turning on your system. This is because the printer device is not present in the hardware graph. To work around this problem, perform the following commands in a shell:

% su

# cd /dev

# ./MAKEDEV

# ioconfig -f /hw

# exit

Full hostname now used by default
For customers installing IRIX on new systems, or reconfiguring existing systems after IRIX is installed or upgraded: the graphical system administration tools now use the fully qualified hostname by default. This affects scripts that assume that only the partial hostname is used. Scripts may need to be modified to use hostname -s rather than hostname. If you wish, you can override this default and use only the partial hostname.
Gang-scheduled graphics processes can cause graphics to hang
Graphics programs using the schedctl(2) system call to choose gang scheduling mode (SGS_GANG) may cause the graphics to hang. (Gang-scheduled processes that do not use graphics are not affected by this problem.)

Origin 2000 Caveats

The following caveats are for large (64+ processors) Origin 2000 systems.

Cumulative for 6.5 through 6.5.11, and still applicable for 6.5.12:

Process Activity Reporter (par) sometimes hangs
Par hangs when the system-monitoring daemon, rtmond, produces a data stream that overloads the clients. As a result, clients that feed par, like padc, will discard events. If the process exited event is discarded by these clients, par hangs. When this happens, you see messages like the following in /var/adm/SYSLOG:
"rtmond[669]: (CPU 0) Client <local>:11:64505 events 1718 dropped (2%) 289 writes(0% push)for 4720560 bytes, 5 push buffers for 80 KB"

As shown in this example, you can tell that par has hung because the message tells you that the dropped count is other than 0.

Before You Install | To Install | Resources | Index