You cannot mix different models of Ethernet cards in the same SpinServer 3300 chassis. Each chassis must have two 3COM Model 3C996-T cards or two Model 3C996B-T cards.

Each SpinServer has two cluster ports. NetApp recommends that you use network switches to connect the cluster links. If you use cables to connect the cluster links directly, the following restrictions apply:

• If the IP addresses of the two cluster links on one SpinServer are on the same subnet, connecting both links with cables is not supported. The only supported case in this scenario is connecting only one link.

• If the IP addresses of the two cluster links on one SpinServer are on different subnets, connecting both links directly with cables is supported.

(Bug 122701)

If your SpinServer uses 3Com® Ethernet card Model 3C996-T, you can experience poor network performance if the ports are directly connected (that is, not through a switch) and the maximum transmission unit (MTU) setting is 9000. To avoid this behavior, NetApp recommends that you connect the ports through a switch. Alternatively, you can set the MTU value to 1500. Note that this model of card is used only in the SpinServer 3300. (Bug 123519)

If you are using a SpinServer 3300, do not connect all four of its HBA ports to the same RAID subsystem. This configuration can result in World Wide Name conflicts. In a supported configuration, up to two connections can be made from a SpinServer 3300 to a single RAID subsystem. (Bug 125022)

When a SpinServer 4100 server is first powered up, the SCSI alarm LED on the front panel begins to blink and the audible alarm sounds. This is due to a faulty temperature reading. To work around this problem, power down and power up the SpinServer. (Bug 125239)

When connecting a SpinServer to storage, be certain to use the Fibre Channel ports that are on the left side as you face the rear of the SpinServer. These ports are listed in the documentation as ports 0 and 1 but are displayed in SpinManager as ports 2 and 3. (Bug 127002)

On a SpinServer 3300, failure or removal of the power supply (PS1 or PS2) does not result in an event notification or an e-mail alarm being issued. (Bug 127387)

Write caching is now enabled by default on the SpinStor 200 RAID enclosure. This can exercise the battery backup unit (BBU) in the enclosure. See the Administration Guide and Administration Reference for information about the software commands that are used to work with the BBU. If the battery in the backup unit fails, it must be replaced. Contact Technical Support for assistance in replacing the battery.

When a RAID set is rebuilt on a SpinStor 100 RAID enclosure, the enclosure beeps. Normally, the beeping ends when the rebuild process finishes; however, if you issue the raid enclosures mute command during the rebuild process, the beeping continues indefinitely after the rebuild process is complete. This is caused by a bug in the RAID firmware used on the SpinStor 100. To avoid this problem, do not issue the raid enclosures mute command during the rebuild process. If the beeping continues after the rebuild process is complete, issue the raid subsystems reset command. This problem does not occur on the SpinStor 200 RAID enclosure. (Bug 127118)

Currently, the only storage products supported with SpinFS 2.5 are the following. See the SpinServer Supported Storage Release Notes for additional details.

• NetApp FAS270c, DS14mk2 FC, and NearStore R200

• NetApp (previously Spinnaker Networks®) SpinStor and SpinStor 200

• Hitachi Data Systems® 9910

• IBM TotalStorage FAStT 900 Storage Server

• StorageTek® D178 Disk Subsystem

When creating a SpinServer storage pool from LUNs on a supported storage product, note that a storage pool striped across multiple LUNs uses the smallest LUN size as the size of all of the LUNs. For instance, if you create a storage pool from three 300-GB LUNs and one 100-GB LUN, the storage pool considers all of the LUNs to be 100 GB, resulting in a storage pool of 400 GB, not 1 TB. NetApp recommends that you make all LUNs the same size so that a storage pool uses all of the storage available on its constituent LUNs.

If a SpinServer's uninterruptible power supply (UPS) cable is disconnected when the system starts up, the SpinFS software does not start up; this is the correct behavior. However, when the cable is reconnected, SpinFS does not start automatically. You must restart the SpinServer after the cable is reconnected to start SpinFS. (Bug 123866)

During the installation of a SpinServer system, make the following configuration changes to optimize performance. Make these changes only to SpinServer systems that use FAS270c storage.

Log in to each SpinServer as root, open the /etc/modules.conf file in a text editor, and add the following lines to the bottom of the file:

options spind lpEnableHolePunching=1
options spinfcd highThrottleLimit=512

Make the same changes in the /altroot/etc/modules.conf file.

Legato® NetWorker® has two methods of doing restoration procedures: the Save Set Recover procedure restores an entire VFS, and a Recover/Directed Recover procedure enables you to choose from a list of directories and files to restore. When you use NetWorker 7.0 for Windows with the NDMP option, a Recover/Directed Recover procedure started from the NetWorker User GUI does not properly initiate a restoration process for a SpinServer. (Note that a Save Set Recover procedure works correctly when started from the GUI.) Instead, this type of restoration procedure must be performed from the command line by using the nsrndmp_recover command. The GUI can be used to browse and select the files that have been backed up; however, when the appropriate directories and files are selected, the restoration of them must be performed from the command line. To restore the directories and files, perform the following steps:

If you use Legato NetWorker 6.1.3 and mistakenly add an incorrect VFS name to the backup list (as shown by a backup failure for that VFS), all the VFSs listed after the incorrect entry cannot be restored, even though NetWorker reports that all VFSs are backed up properly. This is a known issue with NetWorker 6.1.3; the problem has been corrected in NetWorker 7.0. (Bug 125576)

Fully restoring a VFS restores mount points such as .readwrite and .spinshot. When working in a restored VFS, be careful not to perform operations that can potentially cross mount points and affect data. For instance, assume you have data from a backed-up VFS named lisa that you restore under a VFS named ralph. If you issue the rm -rf lisa command in the ralph VFS, the operation can cross the .readwrite mount point and delete all data in the lisa VFS. (Bug 127273)

After you replace a disk drive with a disk drive of a different size, the raid diskdrive show command does not update the size of the disk drive to reflect the size of the new disk drive. (Bug 121896)

If you are using a SpinStor RAID array and have a RAID set that is performing a rebuild operation, the raid sets show command does not display a status of Rebuild in the Long Op In Progress field. Additionally, the raid sets rebuildstatus command does not display the current progress of the rebuild operation. This problem does not occur with the SpinStor 200 RAID array. (Bug 123974)

If you replace a RAID controller on a SpinStor 200 RAID enclosure, you can see the following message:

Lost frame for device -1 [1, 0, -1]

This message can be safely disregarded. (Bug 124899)

The raid commands report the sizes of disk drives in gigabytes; however, the commands interpret “gigabyte” as 2³⁰ (1,073,741,824) bytes, not one billion bytes. Most disk-drive manufacturers interpret “gigabyte” as one billion bytes, so there can be a discrepancy between the output of the commands and the label on the disk. For instance, the commands can report a 36-GB drive as having a physical size of 34 GB. (Bug 127014)

One or more SpinServers in a cluster can stop synchronizing their time settings against external NTP servers. The unsynchronized servers' clocks then run freely and can drift by several seconds per day. To determine whether a SpinServer is not synchronizing its time setting, log in as root and run the following command:

# /usr/sbin/ntpdc -c peers

If the line listing the SpinServer begins with the character x, the SpinServer is affected. To reset the system time, restart the ntpd daemon on the affected SpinServers. (Bug 139511)

Comments added to CIFS shares are not displayed from CIFS clients. (Bug 121046)

Windows 2000 wild-card patterns are not uniformly supported. (Bug 121070)

When the LSA Restrict Anonymous registry key is set to true on the primary domain controller (PDC), CIFS names and SIDs are not resolved. (Bug 121514)

XCOPY32 or ROBOCOPY cannot copy ACLs to a SpinServer that contains access control entries (ACEs) referring to local users (for example, users that are not a member of any CIFS domain). (Bug 121596)

Removing the Archive bit when setting DOS attributes on files is supported on Windows NT® and Windows XP. (Bug 121644)

Windows 95 and Windows 98 remote administration protocol (RAP) calls are not uniformly supported. (Bug 121645)

UNIX symbolic links are not supported under CIFS. (Bug 121735)

On Windows NT 4.0, attempting to disconnect a network drive that is mapped to a SpinServer can become suspended until Windows times out the attempt. (Bug 121825)

When using Microsoft® Internet Explorer browser, the copy and drag-and-drop functions are not supported for moving one VFS to another on the same virtual server. To work around this issue, use one of the following methods:

• Open a Windows command prompt and use the copy command.

• In Windows Explorer, select the file, then right-click it and select Copy. Go to the destination, right-click, and select Paste. Alternatively, select the source file and press Ctrl-C to copy it, then go to the destination and press Ctrl-V to paste it.

(Bug 122432)

When a Windows client writes a file to a SpinServer CIFS share, the client explicitly sets inherited ACLs on the file from the containing folder. (Bug 123617)

If a storage pool goes offline while a file is being copied to it from CIFS, you cannot immediately access the failed copy when the storage pool comes back online. Attempting to access the target file (the failed copy) results in a sharing violation. You must close and reopen the client connection before deleting or otherwise working with the target file. (Bug 123646)

Users or groups that have the same name as the CIFS domain cannot set ACLs. This is believed to be the result of a problem in the Windows Primary Domain Controller (PDC). Note that, by default, Windows does not allow users or groups to have the same name as the domain. (Bug 123862)

Assume you have a VFS mounted at /share and a CIFS share named CIFS_SHARE that points to the VFS. If you delete the VFS, you cannot access the share; this is the correct behavior. However, if you then re-create the VFS and mount it at the same location, you still cannot access the share. (Bug 124085)

If a CIFS user is a member of more than 24 groups, CIFS authentication fails for the user. (Bug 124403)

Only one Kerberos server per Kerberos realm is supported. If you are using Kerberos with CIFS and the primary domain controller (PDC) fails, Kerberos authentication cannot be maintained by backup domain controllers (BDCs). If a Kerberos-enabled PDC fails and authentication is maintained by one or more BDCs, NTLM authentication is used. (Bug 126115)

If you are using Windows 2003 Server with Kerberos authentication, you must add the following Windows Registry keys for Kerberos to work. See the Windows documentation for information about setting Registry keys. (Bug 127145)

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\kdc]
"MaxDatagramReplySize"=dword:00000fa0

[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\kdc\Parameters]
"MaxDatagramReplySize"=dword:00000fa0

When you use Kerberos authentication with CIFS, a SpinServer's system log can fill up with messages of the form streamdaemons: rpcauthkerberosauth_1_svc: Call to krb5_rd_req() failed with error message: Decrypt integrity check failed. To stop these messages, reset the CIFS server's Kerberos password by using the spinaccess cifs servers resetpassword command. (Bug 128539)

You cannot remove a WINS server from a CIFS domain by using the GUI; however, you can remove it by using the CLI. (Bug 128888)

If you restrict the use of CIFS ACLs, CIFS sticky bits are not set because they cannot be determined from NFS mode bits.

If you use Mac OS X (Panther) with Kerberos authentication to access a CIFS-enabled virtual server, the client can report the error Session setup failed (extended security lookup2): syserr = Permission denied, and a CIFS_Auth_Server error is logged on the SpinServer. This problem is caused by a difference in key-caching behavior between the Microsoft and MIT implementations of Kerberos. To work around this problem, use the kdestroy and kinit programs on the client to clear the old Kerberos credentials and get new ones. Alternatively, you can shorten the time-out period for cached Kerberos credentials on the Microsoft Active Directory Controller. To access the caching time-out settings, click Start → Programs → Administrative Tools → Domain Security Policy, then click Security Settings → Account Policies → Kerberos Policy and adjust the Maximum lifetime for user ticket value from 10 hours to 1 hour.

NFS mount requests for root (/) can fail in the time period between creating a mirror for the root VFS and replicating it. (Bug 122909)

In very rare circumstances, a Red Hat Linux client can freeze when attempting to access data on a SpinServer system. The following conditions must be met for this problem to occur:

• There is a high-level directory with multiple VFSs mounted on it, and the VFSs are served by different SpinServers.

• One of the SpinServers fails at the precise moment that you use the client to run the ls command or a similar file-checking operation on the high-level directory.

To work around this problem, restart the client. This problem does not occur with other NFS clients, such as Solaris. (Bug 125459)

If you use Linux clients with a SpinServer system, performance can degrade if the clients are set to use 32k packets over UDP. This is a client-side issue. To avoid this situation, NetApp recommends that you set the clients' socket size to 256k (the default for Red Hat Linux is 64k). For more information about optimizing Linux client performance, see http://nfs.sourceforge.net/nfs-howto/performance.html. (Bug 127206)

The following note applies if you are using Solaris™ 10 clients with a SpinServer system. If a Solaris 10 client has a principal in the form host/client_name instead of root/client_name in its /etc/krb5/krb5.keytab file, the client sends NFS service tickets with host/client_name as the principal of the client. Solaris 10 NFS servers map principals that begin with host/ to the anonymous user (typically nobody). To match this behavior on a SpinServer system, create the following Kerberos-to-NIS mappings by using the spinaccess kerb-nis new command:

(Bug 142719)

If you have a two-node cluster, NetApp recommends that you enable high-availability mode on the cluster by using the cluster configuration enablehamode command. Using HA mode helps ensure that quorum is not lost in the special case of a two-node cluster.

If you are using SpinServers across a WAN and your network drops packets, SpinServer cluster performance can decrease significantly; specifically, this behavior is seen when the maximum transmission unit (MTU) setting is 1500 and the packet drop rate is 1% or greater. To avoid this behavior, ensure that the packet drop rate is below 1%. The severity of this behavior can be substantially lessened if the WAN can support 9000-byte MTUs. (Bug 125310)

If a cluster is modified by using a series of cluster status commands, the cluster can sometimes become unresponsive due to a voting conflict. Commands that can contribution to this state include cluster status addepsilon, cluster status deletenode, cluster status disablenode, cluster status enablenode, and cluster status removeepsilon. To determine whether an unresponsive cluster is in this state, log in as root to the cluster's SpinServers and run the rdb_dump command. The output of this command lists each RDB unit and its state (online or offline). If a majority of the RDB units are listed as online, the cluster typically recovers on its own.

If a majority of RDB units are listed as offline, run the rdb_dump -qf command. If the command output includes the text State: WS_WaitingForVotes, the cluster requires manual intervention. To recover from this state, restart the Tomcat, vldb, and hamgr processes that are listed as Master in the command output. If this does not clear the problem, or if no processes are listed as Master, restart the same processes on SpinServers that are listed first in the output of the rdb_dump -f command.

If you need additional assistance to diagnose or troubleshoot this problem, contact Technical Support. (Bug 135830)

You must issue the storage pools failover command from the SpinServer that is the current primary server for the affected storage pool. For example, assume that a primary server named Server1 fails and Server2 takes over as the primary server. If you want Server1 to resume as the primary server (after it is brought back online), you must log in to Server2 to run the storage pools failover command. The command fails if you run it from Server1. Alternatively, run the storage pools revertpools command. (Bug 122248)

A storage pool can freeze into a passive state if all of the following conditions are met:

• The storage pool is newly configured and no VFSs have been created on it.

• The pool is configured for HA.

• The pool's primary SpinServer and active secondary SpinServer lose access to the pool's storage units, so that the status of the pool is shown as waiting for disks.

Run the tools storage pools forcetakeover command to restore a storage pool that gets into this state. (Bug 143891)

You must have cookies enabled in your web browser to use the SpinManager GUI. (Bug 123773)

The SpinManager GUI can run very slowly if you are using Internet Explorer 6. To work around this issue, perform the following steps in Internet Explorer:

1. Click Tools → Internet Options.

2. Select the General tab. This is the default tab on the Internet Options panel.

3. Click the Settings button in the Temporary Internet Files area.

4. For the Check for newer versions of stored pages setting, select Automatically.

5. Click OK on the Settings panel and the Internet Options panel.

(Bug 124173)

If you are using Netscape® or Mozilla® to navigate the GUI and you select an item from the GUI, then drag and paste it into another window, one or more leading spaces can be added. For instance, if you select the text cd /copyofroot from the GUI in Netscape, then paste it into a UNIX shell window, the text is pasted as cd / copyofroot. This behavior does not occur with Internet Explorer or if you double-click and copy the item before pasting it. (Bug 124840)

If your browser cannot connect to the SpinManager GUI, ensure that all proxy settings on your browser are turned off. (Bug 125531)

As stated in the Administration Guide, NetApp recommends that you use a secure-shell client to access the SpinManager CLI. Secure-shell clients are available from Hummingbird Ltd. and other sources, as well as from the sources listed in the Administration Guide. (Bug 125532)

If you are using Internet Explorer 5.0 or earlier with the SpinManager GUI, the browser hangs if you attempt to access online help. This problem does not occur with Internet Explorer 6.0 or later, or with Netscape or Mozilla. (Bug 127663)

If you are using the CLI to enter multiple arguments for a single parameter, NetApp recommends that you put quotation marks around the arguments to ensure correct parsing. Quotation marks are not required in the GUI. For instance, if you are issuing the event notification handlers new command with values of mail.mycompany.com and mmoore@mycompany.com for the -parameters parameter, enter the command as follows:

spinserver::> event notification handlers new -name foo -type email -parameters 'mail.mycompany.com, mmoore@mycompany.com'

(Bug 127911)

Error messages for the filestorage scheduler jobs commands incorrectly indicate that the commands use the -scheduler parameter, not the correct -schedule parameter. (Bug 127984)

If the SpinServer's hardware clock is not set to Greenwich Mean Time (GMT), you cannot reset the time zone by using the spinservers modify command. (Bug 128882)

If you are logged in to a SpinServer as root and run spinsh commands with the ? parameter to display command parameters, enclose the command and parameter inside single quotation marks, as in the following example:

# spinsh -c 'filestorage vfs show ?'

Not putting the command in quotation marks can have unexpected results. (Bug 139056)

Deleting a mirror's sink does not delete any SpinShots of the sink.

If you have a VFS that is mirrored to two storage pools and you want to delete one of the mirrors, ensure that all mirrors are replicated immediately before deleting any mirrors, or otherwise ensure that all mirrors are synchronized with the source VFS before deleting any mirrors. This prevents clients from receiving potentially out-of-date data from an obsolete mirror. (Bug 126291)

When you are creating a mirror task with the VFS Scheduler, ensure that the VFS specified by the -vfs VFS_name parameter is a source VFS, not a sink VFS (that is, a parent read-write VFS, not another mirror). Specifying a sink VFS as the source of a mirror task can result in errors. (Bug 126453)

If you copy a read-write VFS that has other VFSs mounted below it, the mountpoints of the original VFS are propagated to the copy. For instance, assume you have a VFS named homer that has a VFS named bart mounted below it (/homer/bart). If you make a copy of homer named marge, a mountpoint is created for bart on marge (/marge/bart). NetApp recommends that no more than one mountpoint exist for any given VFS. To delete these superfluous mountpoints, use the tools filestorage vfs deletemountpoint command. (Bug 124757)

Running any of the statistics filesystem vfs commands displays statistics only for VFSs that are directly accessible by the local SpinServer. Statistics for VFSs on storage pools for which the local SpinServer is not a primary server are not shown by these commands. (Bug 127443)

You cannot create a VFS tree with more than 11 mountpoints. (Bug 142405)

SpinFS 2.5 supports 32k read operations over NFS, but SpinFS 2.1 does not. If a cluster includes SpinServers that are running both SpinFS 2.1 and SpinFS 2.5, and a virtual interface fails over from a SpinFS 2.5 system (supporting 32k read operations) to a SpinFS 2.1 system (supporting 8k read operations), a Linux client using the failed-over interface can request 32k, receive only 8k, and appear to have retrieved corrupted data. This is a Linux client problem, not a SpinFS problem. To avoid this problem, configure all virtual interfaces on SpinFS 2.5 systems so that they cannot fail over to SpinFS 2.1 systems. (Bug 127306)

The -floatingsuperuser parameter on the vservers servers modify and vservers servers new commands specifies whether a superuser can log in from any network node. If the value of the -floatingsuperuser parameter is set to false, a superuser can log in only from a specific set of network nodes. If the value of the -floatingsuperuser parameter is set to true, a superuser can log in from CIFS or any form of authenticated NFS (NFS-Kerberos or SpinAuth), on any network node. If you are using unauthenticated (AUTH_SYS) NFS, the -floatingsuperuser parameter has no effect; you must log in from a network node that matches one of the privileged IP address/netmask pairs.

If you use the vservers servers modify command with the -floatingsuperuser parameter on an active virtual server, the command resets all client connections.

If you are restoring data from a VFS dump and the restoration operation is interrupted, the storage pool onto which the data is being restored can be marked BAD and incorrectly list the amount of space used on it. A storage pool that is marked BAD must be fixed by using the SpinFS file-checking utility. Contact Technical Support for assistance. (Bug 126737)