User manual QUANTUM DATAMGR 3.6.2 RELEASE NOTES 09-2002

[. . . ] All messages will be stored in the DataMgr system log file. This information often aids the support organization in determining the exact nature of the problem. After the data is collected, reset the SL_TAC_MASK setting to the previous value and cycle DataMgr. The logging hierarchy has also changed in DataMgr 3. 6. 2. logs log_params (environmental vars) dtrace/ (not used) history/ (cmd history) perf/ (not used) tac/ (system log) trace/ (not used) SSD log The environment variables have changed: · EL_SYSLOG_FACILITY is now SL_SYSLOG_FACILITY · EL_SYSLOG_OPTIONS is now SL_SYSLOG_OPTIONS · EL_SYSLOG_MASK is now SL_SYSLOG_MASK Page 2 of 12 DataMgr 3. 6. 2 Release Notes -- 6-00398-01 Rev A Fixed Problems The following table lists fixed problems: Operating System Solaris 5. 8 Sparc (Solaris 8) HPUX 11. 0 HPUX 11. 0 All Problem Number 09237/ 12944 09517 09812 10061 Problems Fixed in Version 3. 6. 2 xdmfs: Edit Migration Policy GUI does not accept keyboard input. [. . . ] CAUTION: If you have dumped the Trashcan (perhaps to reclaim space on AMASS), or if you have deleted or modified the . version files on AMASS, contact DataMgr technical support so you can resynchronize your BFS Database. On the other hand, if you have not dumped the Trashcan, or if you have never deleted or modified the. version files on AMASS, you can upgrade to this version of DataMgr without first calling technical support as described in the above paragraph. 05211 CAUTION: Before you dump the Trashcan for the first time, you must run the /etc/bfs/usr/utils/bfsdb_bfidck -c utility, which checks for RID reuse in the BFS Database. For option information, refer to the "Utility Reference" chapter in the System Administrator's Guide to Using DataMgr book. The new "ftoken key" field in the BFS Database (described in CR #5407) detects if an AMASS RID (record ID) has been inadvertently associated with multiple bitfiles. (One bitfile ID from the BFS Database should point to a specific RID on AMASS. In turn, this RID points to a specific bitfile under the mount point /archive. Please note that because a BFS can use more than one AMASS file system to archive files, duplicate RIDs are possible. Therefore, DataMgr also checks the SSD name as well as the RID. ) 05407 CAUTION: If you are upgrading DataMgr, make sure you have a successful current backup of your BFS Database. Beginning with DataMgr Version 3. 3. 1, a new "ftoken key" field is created in the BFS Database, and a DataMgr log message will say "Successfully created Bitfile Server database with ftoken key. " CAUTION: Because of this BFS Database field enhancement, do not downgrade to an earlier version of DataMgr without first contacting ADIC technical support. - Upgrade Sequence (Product Bulletin #DM1) If you upgrade your operating system as well as AMASS and DataMgr, perform the following steps: Step 1 Task Make a successful backup of your DataMgr BFS Database. CAUTION: Run the AMASS sysdbchk utility on the AMASS File System Database to make sure there has been no database corruption. 2 3 Make a successful backup of your AMASS File System Database and Journal by running the amassbackup -v command. Upgrade the UNIX operating system. DataMgr 3. 6. 2 Release Notes -- 6-00398-01 Rev A Page 7 of 12 Step 4 Upgrade AMASS. · Task NOTE: IRIX Version 6. 5. x only: This OS version moves the location of the hardware device files from /dev/scsi to /hw/scsi. Therefore, running the AMASS install_tests script returns an error. Consequently, to create the proper links, you must answer the AMASS script questions again about the juke and drive SCSI configurations when you run installamass. 5 Upgrade DataMgr. IRIX 6. 5. 10m Support (Product Bulletin #DM4) DataMgr 3. 6. 1 and later does not support IRIX 6. 5. 10m as stated in the DataMgr 3. 6. 1 Release Notes. An IRIX bug (841143) was found that keeps DataMgr from starting. ADIC recommends that DataMgr 3. 6. 1 or later be kept at 6. 5. 9m or below. Operating Issues The following issues apply to this release: Operating System All CR Product Alert #DM48 03959 Description CAUTION: Do not remove any files under /archive/FMSclients on the AMASS server. Deleting or modifying these files will corrupt the BFS Database. The following messages indicate a problem first reported in SGI bug 363715 that currently does not have a patch: · · IRIX 6. 2 "WARNING: second try at anon pfind failed, vaddr 0x4000000. " "WARNING: Process [ssd_amass_nn] pid 10021 killed due to bad page read (errno 2). " Workaround: Reboot the server to fix what is on swap and report the problem to SGI. All #4318 Product Alert #DM55 06411 Using the UNIX cpio command while DataMgr is migrating the same file sends cpio into an infinite loop. Workaround: Use the UNIX tar command to back up files. [. . . ] This problem has been fixed. DataMgr 3. 6. 2 Release Notes -- 6-00398-01 Rev A Page 13 of 12 Problems Fixed in Version 3. 6. 1 The following table lists fixed problems: Operating System HP9000/800 10. 20 Problem Number 08548 Problems Fixed in Version 3. 6. 1 Systems hang after moving many small files to migration partition using the mv command. NOTE: 128 MB of memory is suggested on HPUX systems. For anything less than 128 MB, use "dmfscntl -c 15000" to drop the maximum number of kernel requests queued in the dmfs driver from 50000 to 15000. ALL ALL 08902 09389 Have SLD recover missing registered file on system startup Add option to dmfstat to print the different states of a file. The following options are new: -b files that in the process of being migrated -e files that are currently stubbed -h files that are currently shadowed -o files locked from migration -r files that are currently eligable to be migrated -x files that were added by dmfsaudit SGI_Origin_2000 6. 4. 1 HPUX 11. 0 IRIX 6. 5. 3 Solaris 5. 8 Sparc (Solaris 8) Solaris 5. 7 (Solaris 2. 7) 09531 09690 09698 09917 09943 Add dmin messages to transaction log files so reads from AMASS can be identified HP-UX system processors will hang with large negative spinlock values Spaces in file names in locklist can cause migration of certain files to stop dm_snap falsely reports that DataMgr is not installed After an upgrade from SGI to Sun, there is a mismatch between the expected and actual size of the file, so DataMgr won't delete the files automatically. A new option, -b, was added to bfsdb_bfidck which will fix the blocksize found in the bfs database if it is half the number of blocks of the file in Amass. HPUX 11. 0 Solaris 5. 7 (Solaris 2. 7) Solaris 5. 7 (Solaris 2. 7) Solaris 5. 8 Sparc (Solaris 8) AIX 4. 3. 3 10072 10085 10192 10405 10410 dmfscntl -l, with a large locklist, can take hours to finish bfsdb_bfidck should display a useful message when policy is missing for file instead of just exiting bfsdb_bfidck can core dump when it finds an error in the migration file system Add license info to dm_snap so expired license string can be easily identified Excessive amount of NFS errors on AIX platform when doing large number of file transfers. Page 14 of 12 DataMgr 3. 6. 2 Release Notes -- 6-00398-01 Rev A Operating System IRIX 6. 4. 1 Problem Number 10633 Problems Fixed in Version 3. 6. 1 install_DM shouldn't overwrite exisiting startup files during upgrade Previous DataMgr Releases For information about previous features and fixes to the base software code, refer to prior Release Notes, which can be found on the ADIC secure site. [. . . ]