1. Fault description The complete EVA storage structure is composed of one EVA4400 controller, 3 EVA

Operating technique: HP-UXB.11.31 Database: Oracle10.two.0.1.0 Enterprise Edition-64bit for HPUX

The following is often a brief introduction for the installation atmosphere: (1) Atmosphere detection # uname-allHP-UXbyhpux1B.11.31Uia641447541358unlimited-userlicense The machine is IA64 architecture, the operating system is HP-UX, as well as the version is B.11.31. Then verify the storage space information of every single component to make sure that there's adequate space. (2) Detection and installation of dependent packages Check the patch packages expected by oracle10g in accordance with the installation instructions 'b19068.pdf'. Detection: # swlist-lbundle | grep 'GOLD' # swlist-lpatch | grepPHNE_31097 If nothing is detected, you must visit the official web site to download and set up it. Set up the patch package: swinstall-s / patchCD / GOLDQPK11i-xautoreboot = true-xpatch_match_target = true (3) develop customers and groups # groupadddba # useradd-gdba-d / house / oracleoracle # passwdoracle (4) make directories and modify permissions to create directories : # Mkdir–p / opt / oracle / solution / 10.2 / oracledb # chown-Roracle: dba / opt / oracle / frombyte.com modify permissions: #chownoracle: dba / usr / oracle_inst / database / # chmod755 / usr / oracle_inst / database / (5) Set atmosphere variables vi / residence / oracle / .profile (six) Install Oracle Oracle installation needs a graphical interface, so initial test that the graphical interface can start off ordinarily. # exoprtDISPLAY = 192.168.0.1.0: 0 $. / runInstaller Immediately after the graphical interface is installed, the installation is comparatively straightforward. Here, only the software is installed, and no examples are installed. (7) Test the database connection # su-oracle $ sqlplus / assyssdba4, confirm the Oracle database (1) Mount the raw device Considering that a few of the LUNs are raw devices, the LUNs we recover are all within the kind of files. Therefore, you'll want to mount the file LUN on HP-UX. Install the iSCSIInitiator on the HP-UX server, the installation methods are as follows: verify irrespective of whether the computer software package is total #swlist-d@/tmp/B.11.31.03d_iSCSI-00_B.11.31.03d_HP-UX_B.11.31_IA_PA.depot install the software program package # swinstall- xautoreboot = true-s / tmp / B.11.31.03d_iSCSI-00_B.11.31.03d_HP-UX_B.11.31_IA_PA.depotiSCSI-00 add the iSCSI executable file to PATH # PATH = $ PATH: / opt / iscsi / bin / frombyte.com detects regardless of whether iSCSI is installed effectively # iscsiutil-lConfigure the initiator name of iSCSI # iscsituil / dev / iscsi-i-Niqn.2014-10-15: LUN configuration mount target iSCSI device # iscsiutil-a-I10.10.1 .9 Delete the target iscsi device # iscsiutil-d-I10.10.1.9 Verify that the target iSCSI is mounted effectively # iscsiutil-pD Discover the target target device # / usr / sbin / ioscan-H255 Create a device file for the target # / usr / sbin / insf-H255 (2) Import external VG facts to make a VG node # mkdir / dev / vgscope / frombyte.com Produce a VG device file name # mknod / dev / vgscope / groupc640x030000 Check whether or not the PV is standard # pvdisplay-l / dev / dsk / c2t0d0 / frombyte.com import PV into VG # vgimport-v / dev / vgscope / dev / dsk / c2t0d0 activate VG info # vgchange-ayvgscope view VG details # vgdisplay-vvgscope

Figure six: (3) Modify the LV name simply because it really is a VG rebuilt on a new environment, after which import the PV in to the newly produced VG. For that reason, the names of LVs have all been changed. You should manually modify the names of LVs to the ones prior to. Figure 7: Mainly because there are actually two original database instances, along with the raw device storage is applied. Consequently, when developing a database instance, it really should be configured and named in accordance with the original. At the file technique level, with simultaneous help, all LVs are mounted and permissions are modified. rescue ntfs : Instance of installing the database. In accordance with the original configuration, together with the help from the customer DBA, all bare device files are installed and identified. Then adjust the configuration parameters, verify the database storage status, and prepare for beginning the database. Initially switch to the instance scope (most important). To start the database. SQLstartupmount; SQLselectfile #, errorfromv $ recover_file;-check the damaged file. There is no damaged file. SQLALTERDATABASEOPEN; no error is reported at startup, nevertheless it is slow, then the user is queried, and two tables of a user are queried randomly, and also the data outcome set returns to typical. Then the connection was abruptly interrupted, reconnected, and also the check status was that the database was closed. In the event you start off the database once again, it will nevertheless fail to begin and will be forced to close. After preliminary testing and routine restoration in the library state, this issue can't be fixed. Verify that the NJYY database switches the environment variable to another database, NJYY. When opening the database, it reports an insufficient memory error and can't open the database. Following preliminary inspection, the information file was not broken. SQLstartupmount; SQLselectfile #, errorfromv $ recover_file; SQLALTERDATABASEOPEN; error4030detectedinbackgroundprocess5, repair Oracle database fault repair For the scope database, based on the above operations and fault phenomena, the initial judgment is that there is certainly a problem using the undo table space or log. Check the integrity and consistency of your information file, only one particular undo01.dbf file is damaged. It is actually determined that the undo tablespace is broken. Delete the broken undo tablespace via the command and rebuild it in its original place. Verify other components of the file and located no problems. Restart the database, start out usually, query information, regular, integrity verify, normal. Then do the entire database export of imp database. Just after additional than three hours, export the entire database usually. For the NJYY database, it is detected that the memory space setting is incorrect. Right after the command adjustment, the database returns to regular, and may be began normally and made use of usually. Lastly, export the whole database from the imp database. Right after more than four hours, export the whole database normally. Particular verification Immediately after completing the initial verification, Celebration A needs its DBA and small business personnel to do further specific verification by means of a remote database. In conjunction with the verification with the verification atmosphere and each database. Lastly verify that the database is fully restored with out difficulties. Just after verifying the data, do the data migration. Contemplate the database capacity and recovery time. Pick to use expdp to export the entire database information. Since expdp is much more efficient than exp. Following writing the export script and testing with out troubles within the test atmosphere, very first export the scope database. 24 minutes after the export began, an error began to become reported: ORA-39171: Jobisexperiencingaresumablewait.ORA-01654: unabletoextendindexSYSTEM.SYS_MTABLE_00003A964_IND_1by8intablespaceSYSTEM Right after discovering the cause, it was identified that the method table space was full. Exporting with expdp will add export record information to the Technique.SYS_MTABLE_00003A964_IND_1 table inside the system table space. When exporting a large amount of data, the data volume of this table will increase, and when the total capacity in the method table space is reached, an error will be reported . In get ntfs data back with the analysis right here, the table space generally automatically increases the capacity, so there must be no error. Lastly, it was queried that the system table space was placed around the bare device, the capacity was 1G, and it could not be increased. Therefore, you can't use the expdp tool for export. Can only be exported using the exp tool, even though it will be slower, but there will likely be no issue of insufficient method table space. Ultimately, the entire library is exported to the scope by means of exp, as well as the backup is successfully completed right after much more than 6 hours. The backup file reaches 172G. For the NJYY database, do imp export. Just after much more than 7 hours, the whole database is exported normally, and also the backup file reaches 140G. Then the regional backup from the database backup file is made as a secure cold backup. V. Transfer the data 1. Right after transferring the vmware virtual machine file and Oracledump file to confirm that all information is appropriate, copy the vmware virtual machine file and Oracledump file to a 2TB Seagate really hard drive. Then copy the recovered LUN information to two 3TB single disks. Following coming for the internet site of Celebration A, initial hand more than the vmware virtual machine file and Oracledump file to Party A, then Celebration A begins to confirm the dump file and vmware virtual machine file. 2. Mirror the LUN information to Celebration A's EVA4400 storage server. Because Party A demands all LUN data to be restored to the original atmosphere, HP-EVA4400 must be reconfigured to recreate the same size LUN as ahead of. Then use the winhex tool to mirror each of the recovered LUN data for the newly designed LUN of EVA. As a consequence of some troubles with all the HP-EVA4400 controller of Party A, it took a lengthy time for you to debug and reset the HP-EVA4400. Immediately after mirroring all LUN data, Party A arranges for an Oracle database engineer to verify that the restored Oracle is regular. Following the detection, it was identified that two dbf files have been lost, which caused the Oracle service to fail to begin. Right after analyzing the lead to in the failure, it was identified that these two lost dbf existed as files just before the EVA failure, and later restored to LV when recovering Go inside. The Party A engineer didn't rebuild the vg when restoring the LV, but restored all the LV in accordance with the preceding vg_map. Consequently, this trouble only occurs. The option is to re-create the two LVs, and after that take away the two files in the underlying LUN and dd them to the newly developed LV. Start off the Oracle service once again, the startup is normal, plus the issue is solved. Due to the fact the on-site atmosphere is well preserved after the failure, no relevant unsafe operations are accomplished, which can be of wonderful aid to the later information recovery. Even though numerous technical bottlenecks had been encountered for the duration of the whole data recovery approach, they were all solved one particular by one particular. Ultimately, the whole information recovery was completed within the anticipated time, and the recovered information was also very satisfactory. Ideas for future information security 1. Arrange employees to pay a visit to the laptop or computer area frequently and learn if there is any alarm information. 2. Management personnel really should be cautious when handling storage to prevent information loss because of misuse. 3. Some modules on the EVA controller are identified to become unstable around the spot and must be replaced in time. 4. For the reason that the EVA storage failure is caused by the instability from the disk, this a part of the disk need to be the same batch of disks. Hence, the efficiency of these disks is also reaching the limit. If achievable, it's suggested to replace these disks.