POLYCENTER NetWorker Save and Restore for Digital UNIX Version 4.2B Release Notes The following release notes for NetWorker Version 4.2B describe enhancements, features, and restrictions for this release. These release notes are included in your distribution kit. The PostScript file is titled brx42b_relnotes.ps and the text file is brx42b_relnotes.txt. Copies of the release notes are included in the Release Notes and Documentation subset and are located in the directory /usr/opt/BRX422/usr/doc after you install this subset. The release notes are included in this location to ensure that they are accessible from the product directory after the distribution disk is unmounted. See the Installation Guide Version 4.2A (in the brx42b_iguide.ps file) for instructions on how to install this subset. The installation information is still accurate for this version, except that you should substitute brx42b for every instance of brx42a, and substitute 422 for every instance of 421. Mandatory Update If you are running NetWorker on a system running Version 4.0 (or higher) of the Digital UNIX operating system, you must upgrade to at least NetWorker Version 3.2A. Previous versions of NetWorker are incompatible with Version 4.0 (or higher) of Digital UNIX. This is a requirement for both NetWorker servers and NetWorker clients. If you are running NetWorker Version 4.2A, it is highly recommended that you upgrade to NetWorker Version 4.2B. If you are upgrading from NetWorker Version 3.2A or earlier, upgrade directly to NetWorker Version 4.2B. Do not upgrade to NetWorker Version 4.2A first, because there is a problem which can cause corruption of very large indexes during the installation procedure. Licensing Starting with NetWorker Version 4.2A, use of a sequential loader/stacker device requires the purchase of a Tier-1 Jukebox [C] Digital Equipment Corporation 1996. All rights reserved. license. If you use a sequential loader and are upgrading from NetWorker Version 3.2A or earlier, do not upgrade to NetWorker V4.2B until you have installed this license, or you will not be able to use your loader. You will also not be able to easily revert to using an earlier version of NetWorker, because the installation procedure for NetWorker V4.2B converts NetWorker Version 3.2A and earlier server and client index files to a new format. See the Software Product Description (SPD) for part numbers and ordering information. Refer to the NetWorker Installation Guide for complete descriptions of each license and information on how to install the licenses. The following information may aid in understanding NetWorker licensing policies: o Client, jukebox, and cluster client licenses are assigned to the defined resources in the order in which the server reads them from the resource database (/nsr/res/nsr.res) at NetWorker startup. o Defined clients that are unable to obtain a connection license at server startup will be allowed to browse and recover previously saved data, but will be unable to perform any backups. o Cluster hosts and cluster services do not require an archive license. The ability to perform archive operations is included in the cluster client license. o For tiered licenses (jukebox and cluster client licenses), if no units of the correct tier license are available for use, the next higher tier license will be used, if available. This continues until all available tiers are searched. In addition, a PC client can use a workstation connection license if there are no units of PC licenses available. Enhancements and New Features The following new features and enhancements were first included in NetWorker Version 4.2A. NetWorker Version 4.2B is a maintenance release and does not include any enhancements or new features. o CDE Integration -- This release of NetWorker includes support for launching NetWorker from the Common Desktop Environment (CDE). The NetWorker icon will display in the CDE Application Manager area of the desktop. To start the NetWorker GUI from CDE, double-click the NetWorker icon. o New Authentication/Authorization Scheme -- This release of NetWorker uses name-based authentication. Previously, accounts with matching UIDs on client and server were required for some operations. With this enhancement, reserved 2 NetWorker Release Notes ports are no longer required, except for compatibility with older NetWorker clients. o Improved Server/Client Backup Services -- The nsrexecd process provides a secure and restricted interface through which a NetWorker server can provide backup services to a client. With this release of NetWorker you can establish a list of servers with permission to access the machine for each client and server running NetWorker Version 4.2A or higher. The list of servers is contained in the /nsr/res/servers file, which is created in one of the following ways: _ On a NetWorker client, the installation procedure, nsr_ize, will prompt you for server names and then create the /nsr/res/servers file. To change or add servers later, edit this file. _ On a NetWorker server that you want to act as a client to another server, run the script /nsr/bin/nsrexecd.sh following completion of the NetWorker installation or upgrade. The script will create the /nsr/res/servers file and can be used later to add other servers to the file. For more information, refer to the nsrexecd(8) reference page. o Save Performance Improvements -- Agent nsrds were removed for this release of NetWorker, which improves performance by eliminating four kernel copy operations. In addition, performance is improved by using a more efficient save record format. Agent saves and the old record format are still supported for compatibility with older clients. o Index Performance and Functional Improvements -- With this release of NetWorker, the index check performed at NetWorker startup is much faster and is performed in place, which results in improved index performance. In addition, index error logging has been improved and a new diagnostic tool, nsrinfo, is available. o Pre/Post Processing for Saves -- With this NetWorker release, a backup command script can be specified for each client resource. The script can contain a set of commands to be executed before starting the client save, as well as a set of commands to execute following the save. The script must be stored in the same directory as the NetWorker save command, and its name must begin with either save or nsr. For example: /usr/bin/save-command.sh: #! /bin/sh PATH=/sbin:/usr/sbin:/usr/bin: date > /tmp/pre-save.date save $* date > /tmp/post-save.date NetWorker Release Notes 3 o Smart Media -- This release of NetWorker includes the following features that automate media and jukebox management: _ Auto Media Management --NetWorker will automatically recycle media that is expired or has been marked auto recyclable in a jukebox with auto media management enabled. It will also automatically label volumes that do not have NetWorker labels when volumes are needed for a save operation. Care must be exercised if volumes without NetWorker labels will be stored in a jukebox with Auto Media Management enabled. _ Automated Cleaning Cartridge Support -- NetWorker allows one or more jukebox slots to be designated to hold cleaning cartridges. If the auto clean field in the jukebox resource is selected, jukebox devices will be cleaned on an established schedule. _ Auto-Media Verify -- Auto-media verification is a new option for media pools. During a save operation, the last successfully written record is held in a buffer and, when the volume is full, the last record is read from the volume and compared to the contents of the buffer. If the compare fails, all saves in progress are terminated. o Save and Recover of AdvFS Fileset Quota and Quota Files -- This release of NetWorker provides support for save and recover of AdvFS quota files for Digital UNIX Version 4.0 and above. Earlier versions of NetWorker did not back up AdvFS quotas. o SNMP Notification -- This new feature sends event and trap notifications to an SNMP-compliant network management station. o New SSC Environment Variable -- A new environment variable for users with Save Set Consolidation (SSC) licenses is provided in this release of NetWorker. The new DIGITAL_SSC_NBUFS environment variable specifies to savegrp the number of common buffers for ssc to use. If no value is provided for this environment variable, savegrp will determine a value based, very conservatively, on the total memory of the CPU. By setting a value for this variable you can potentially improve performance, depending upon your system configuration. The value assigned to this environment variable must meet the minimum and maximum requirements for the ssc command -B switch. For example, to automatically use 32 common buffers, set the environment variables using the following command: # setenv DIGITAL_SSC_NBUFS 32 This environment variable must be set before nsrd is started. See the ssc(8) reference page under the -B flag for details on how this switch affects ssc. 4 NetWorker Release Notes Problems Resolved Since the Last Release The following problems and restrictions have been resolved in NetWorker Version 4.2B: o The NetWorker Version 4.2A installation procedure corrupted indexes larger than 2 gigabytes during index conversion. Also, NetWorker Version 4.2A index files, i.e. /nsr/index/clientname/db, that grow larger than 2 gigabytes may become corrupted. A side effect is that NetWorker Version 4.2A may report strange or negative values for size, usage, or utilization of indexes. o NetWorker Version 4.2A contained a memory leak that occurred when saving indexes. The memory leak resulted in error messages similar to the following: * piedpiper.pa.dec.com:index 2 retries attempted * piedpiper.pa.dec.com:index nsrindexasm: save failed on /a2/nsr/index/piedpiper.pa.dec.com/db * piedpiper.pa.dec.com:index save: external ASM `nsrindexasm' exited with code 1 * piedpiper.pa.dec.com:index save: /a2/nsr/index/piedpiper.pa.dec.com/db was not successfully saved * piedpiper.pa.dec.com:index save: SYSTEM error, I/O error * piedpiper.pa.dec.com:index save: save of /a2/nsr/index/piedpiper.pa.dec.com to devnull.pa.dec.com failed In severe cases this memory leak could cause a server system to hang if NetWorker allocated all available swap space. o In the client definition of a cluster service, when All was specified in the Save Set field, NetWorker Version 4.2A only backed up one of the filesystems in the list. o In NetWorker Version 4.2A, it was not possible to back up clone filesystems of AdvFS cluster services. o When a new client was defined using NetWorker Version 4.2A, the server asked for an archive license regardless of whether archive services had been selected for the client. The license was not permanently allocated, but if no extra licenses were available, the client was not defined. o NetWorker Version 4.2A and earlier versions of NetWorker were unable to verify or clone an archive with a path length greater than 64 characters. The following problems and restrictions were resolved in NetWorker Version 4.2A: o The NetWorker client subset can be installed without an active client license. A message indicating that there is no active client license will display but the installation will complete. NetWorker Release Notes 5 o The nsrck command no longer core dumps if there is inadequate disk space on the server system to hold the client index temporary file while nsrck corrects inconsistencies in a client index. o The available slots attribute for jukeboxes described in the NetWorker Administrator's Guide now functions properly in this release. o Starting a jukebox inventory operation using the nwadmin interface and then canceling the operation stops the inventory. o Multiple changes to the groups resource in the nwadmin program no longer cause nwadmin to core dump. o You can restrict the devices for a clone pool. Attempts to restrict the devices for a clone pool no longer results in an error message. o There are two ways for a NetWorker server to initiate an archive operation on its own files: from the nwarchive interface or from the nwadmin interface. With this release, there are no problems using either interface to initiate an archive request for the server's files. o After an archive operation, an attempt to retrieve files from the archive client will not result in an error message. o The savegroup -p command no longer leaves a group with a status of Running. o Client indexes will not be inadvertently removed by the reclaim space function. Known Problems and Restrictions in This Release The following sections contain information on known problems and restrictions in this release. Workarounds are provided where available. In addition to the information presented here, see Chapter 7, Troubleshooting, in the NetWorker Addendum for further information on configuring and operating NetWorker. Devices and Media o Changing a Device From Sequential Loader Mode to Random Jukebox Mode -- When changing a jukebox device from sequential loader/stacker mode to random jukebox mode, you must first delete the loader resource and then create the jukebox resource. After completing these steps, if no more loader resources are present, savegroups will display the following series of messages when a volume request is made: 6 NetWorker Release Notes 9/19/96 16:09:08 nsrd: nsrld info: nsrld active... nsrld active... could not find a loader resource nsrld exit... 9/19/96 16:09:09 nsrld: nsrld info: could not find a loader resource 9/19/96 16:09:09 nsrld: nsrld info: nsrld exit... These messages appear because the loader notification was not deleted when the last loader was removed. This problem can be resolved by deleting the Loader request notification in the nwadmin interface. These messages have no effect on successful completion of the savegroup. o Stand-alone Device and Auto Media Management -- If Auto Media Management is enabled on a stand-alone device and a request for a tape for backup is pending, NetWorker will take exactly 1 hour to attempt to mount or relabel (as appropriate) the tape loaded on the device. After the tape is mounted or relabeled, the server operations on that device continue normally. If the tape in the device needs to be mounted, you can avoid the 1 hour waiting period by temporarily disabling Auto Media Management on the device and then mounting the tape manually. If the tape loaded in the device is not a NetWorker tape, the tape is labeled without any waiting period. o Labeling a Volume -- If you use the nwadmin GUI to label a volume in a stand-alone device, the label template for the selected pool will not be incremented. You must select the Label Template window from the Customize pull-down menu and manually increment the label template following the label operation. To avoid this problem, label volumes in stand- alone devices using the nsrmm command. o Upgrading from NetWorker Version 3.0 -- If the server is currently using one or more of the jukeboxes and media types listed in the following tables, you must remove the devices and the jukeboxes from your NetWorker configuration before installing this version of NetWorker. Be sure to unmount any loaded volumes before deleting a device or jukebox. (Not doing so will cause an error that requires manual intervention.) After installing this version, create the new device and jukebox resources using the following attributes, then run an inventory for the newly created jukebox. NetWorker Release Notes 7 Media Version 3.0 Version 4.2B 4mm 5 GB 4mm 4 GB 4mm 8 GB 4mm 8 GB Jukeboxes Version 3.0 Version 4.2B ATL ATL/Odetics SCSI Lago Datawheel STK-9708/Lago380 (SCSI) Datawheel TZ857, TZ867, TZ875, Quantum DLT/Digital DTL TZ877 TLZ6L, TLZ6L-12 Digital 4mm DAT (TLZxx) o Selecting a Device When Defining Pools -- The device field in the pool resource does not completely limit use of devices as documented in the nsr_pool(5) reference page. Do not attempt to manually label and mount devices not selected for a pool. The operations will not complete successfully, and you may or may not receive an error message (depending on whether the operation was initiated from the command line or the GUI). o Labeling Two Volumes Concurrently -- Do not attempt to label two volumes concurrently if both label operations use the same label template. Jukeboxes o Canceling a Jukebox Label Operation -- If you start a label operation using the nwadmin interface and then cancel the operation, NetWorker might not move the volume from the jukebox tape drive back into the volume's original slot. 8 NetWorker Release Notes Attempts to label another volume will result in the following message: There is another volume in the drive If this problem occurs, reset the jukebox or issue the nsrjb -u command to reload the jukebox tape drive. o Resetting a Jukebox -- For a jukebox with a tape that is mounted, issuing the nsrjb -H command may not properly reset the jukebox. The jukebox drive will be unloaded but the NetWorker Administrator window and the nsrmm command will continue to display the device as mounted. To avoid this problem, issue the nsrjb -u command to unmount the volume before issuing the nsrjb -H command. If the jukebox gets into the situation described previously, use the nsrmm -u command to unmount the tape. The jukeboxes affected are the optical, DAT (TLZxx), and DLT (TZ8xx). o Resetting a TL8xx Jukebox -- For a TL8xx jukebox with a tape that is loaded or mounted, issuing the nsrjb -H command may not properly reset the jukebox. The jukebox drive will not be unloaded (unmounted) and the nsrjb -H command will return with the following error message: nsrjb: Jukebox error, Illegal Request, Medium Not Present. To avoid this problem, issue the nsrjb -u command to unload (unmount) the volume before issuing the nsrjb -H command. o Sequence Number Mismatch Error When Updating a Jukebox Resource -- If the jukebox resource has been updated since the data was fetched for display in the Jukebox window, an attempt to delete a jukebox results in a sequence number mismatch error. Close the Jukebox window, then re-open it to display current information and allow normal operations. Graphical User Interface o Retrieve Window Problem -- In the Retrieve window, the online help Cancel button might not work unless all other dialog boxes have been closed. Cloning o Incorrect Message for Automatic Cloning of Save Sets with a Level of Skip -- During an automatic cloning operation of a save set with a level of skip, a message similar to the following is displayed: NetWorker savegroup: (notice) xyz_group completed, 4 clients (Save Set Cloning Failed) Start time: Wed Feb 8 20:33:00 1996 End time: Wed Feb 8 20:33:12 1996 NetWorker Release Notes 9 Automatic cloning of save sets to pool abc_pool was never started. In this situation, the Save Set Cloning Failed message is misleading. Because of the skip level, the cloning operation is not started. o Automatic Savegroup Cloning with 4mm Devices -- In some circumstances, a 4mm device can hang at the start of an automatic clone operation. If this problem occurs, see the comments in the /sbin/rc3.d/S94nsrd file. Follow the instructions for a workaround and then restart NetWorker using the /sbin/rc3.d/S94nsrd start command. Do not make this change unless there is a known problem with this type of device. Firmware version 4BQH prevents this problem with the TLZ7L device. o Cloning Save Sets Using Jukeboxes -- For a server with two jukeboxes that each have only one device, NetWorker might hang while attempting to clone save sets that cross volumes from one jukebox to another. To avoid this situation, do not attempt to clone these save sets. Another way to avoid this situation is to label all the volumes in one of the jukeboxes for cloning pools only. If you want to clone some of the save sets on a volume, use the following command to determine which save sets are completely contained on that volume: # mminfo -avo t vol_name The output from this command provides the save set ID number for each save set. In addition, the fl column indicates whether only part of the save set is contained on the volume. The ch, cm, and ct columns designate the head, middle, or tail of the save set. Once you decide which save sets to clone, create a file containing the save set IDs of the save sets you want to clone, excluding the save sets that cross to a volume in the other jukebox. In the file, list one save set ID per line, leaving no blank lines. Make sure there is a carriage return after the last entry. Then, issue the following command to clone the save sets (clone_pool is the name of an existing clone pool, and ids_file is the name of the file you created containing the save set IDs): # nsrclone -b clone_pool -V -f ids_file Miscellaneous o Incorrect Client Machine Type for Windows NT Clients -- Versions of the NetWorker for Digital UNIX server up to and including NetWorker Version 4.2B incorrectly define newly created NetWorker for Windows NT Version 4.3 clients as type Workstation rather than type PC. On NetWorker Version 4.2A or 10 NetWorker Release Notes 4.2B, this causes a workstation license rather than a PC license to be allocated for the client. The workaround is, on the NetWorker server, manually set the machine type field of the NT client resource to PC. This causes the correct license type to be allocated. o Sparse Files -- NetWorker determines that files are sparse (or "holey") by comparing the allocated blocks with the byte size (this can be seen using the ls -ls command). If the allocated blocks do not account for the size of the file, the file is considered to be sparse and saved using an algorithm that replaces long strings of zeroes with "holes" in the recovered file. When AdvFS clone filesets are used, AdvFS reports the number of blocks allocated to the clone (which is zero, or the number of blocks on the real file that have been modified since the clone creation). Thus, files in a cloned fileset always look sparse to NetWorker. The problem that may occur is that some files that were not sparse when saved may be sparse when recovered. Note that Oracle databases are zero-filled, fully allocated files and, as such, are not "holey." They are particularly susceptible to this problem. At this time, the workaround for this is to use the cp command to copy the file after recovery. This will cause a sparse file to be converted to a fully allocated file. A more permanent solution is being implemented by AdvFS in a future release of Digital UNIX. o Retention Policies Greater than 26 Years -- Retention policies greater than 26 years do not work properly. If, going backward in time, the duration represented by the retention policy takes you prior to January 1, 1970, the volumes will be incorrectly marked recyclable. o Head and Middle Volumes Not Marked Recyclable -- If multiple volumes are used for a single save set and one of those volumes is deleted, all the other volumes should be marked recyclable. Currently the volumes are not being marked recyclable by NetWorker, so the volumes must be marked recyclable manually. o Multiple Client Creation -- If you are creating multiple clients and some clients are not created due to license failures, the list of existing clients is not updated with the successful additions until the window is closed and reopened. o Recover Operation Error Message -- When you issue the recover command as an ordinary user (a user other than root), you might see the following message: quotactl: invalid argument NetWorker Release Notes 11 This message is invalid and can be ignored. o Dataless Management Services (DMS) Clients -- NetWorker does not support DMS clients as NetWorker servers. A DMS client can, however, be a NetWorker client. For instructions on how to install the NetWorker client software on a DMS client, contact the Customer Support Center (CSC). o Specifying /proc in a Directive -- Specifying /proc in a directive can result in all save sets failing. For example, do not set up a directive to skip the /proc file system during backups. o Backup with an AdvFS Root File System -- If you are backing up a client running Digital UNIX 3.2x that has an AdvFS root file system, you may encounter an AdvFS bug that causes NetWorker to fail to properly back up mount points on root. If you encounter this problem, contact the CSC for a patch to the AdvFS file system. o Backup of AdvFS File Systems for Digital UNIX Version 3.2C, 3.2D, and 3.2F -- If you are backing up a client running Digital UNIX Version 3.2C, 3.2D, or 3.2F, the AdvFS file system may not be saved properly. You can resolve this AdvFS problem by either installing the consolidated AdvFS patch that corresponds to the version of Digital UNIX you are running or upgrading your Digital UNIX system to Version 3.2G. Contact the CSC for information about how to obtain these patches. 12 NetWorker Release Notes