Improve RepServer performance

— Various parameters that can improve the performance of the Replication server rs_configure memory_limit, ‘200’ — dsi_sqt_cache_size alter connection to DCOCCDWP.defacto set dsi_sqt_max_cache_size to ‘62914560’ go suspend connection to DCOCCDWP.defacto go resume connection to DCOCCDWP.defacto go — Cache system tables configure replication server set sts_full_cache_rs_classes to ‘on’ configure replication server set sts_full_cache_rs_databases to ‘on’ configure replication server set sts_full_cache_rs_columns to ‘on’ configure replication server set sts_full_cache_rs_config to ‘on’ configure replication server set sts_full_cache_rs_datatype to ‘on’ configure replication server set sts_full_cache_rs_functions to ‘on’ configure replication server set sts_full_cache_rs_objects to ‘on’ configure replication server set sts_full_cache_rs_publications to ‘on’ configure replication server set sts_full_cache_rs_queues to ‘on’ configure replication server set sts_full_cache_rs_repdbs to ‘on’ configure replication server set sts_full_cache_rs_repobjs to ‘on’ configure replication server set sts_full_cache_rs_diskaffinity to ‘on’ configure replication server set sts_full_cache_rs_routes to ‘on’ configure replication server set sts_full_cache_rs_systext to ‘off’ configure replication server set sts_full_cache_rs_sites to ‘on’ configure replication server set sts_full_cache_rs_version to ‘on’ configure replication server set sts_full_cache_rs_users to ‘on’ configure replication server set sts_full_cache_rs_translation to ‘on’ go rs_configure sts_cachesize, ‘3500’ go /* REP15 extra settings: — Parallel dsi alter connection to DCOCCDWP.defacto set parallel_dsi to ‘on’ go suspend connection to DCOCCDWP.defacto go resume connection to DCOCCDWP.defacto go rs_configure exec_sqm_write_request_limit, ‘983040’ go rs_configure md_sqm_write_request_limit, ‘983040’ go rs_configure smp_enable, ‘on’ go */ — Also a good idea to use disk affinity to bind connections to their own partitions: — e.g alter connection to DCO.ProductCatalog_new set disk_affinity to ‘P08’

Sybase ASE upgrade process

1.       First you need to create a new directory on Unix for the installation, e.g /progs/fps/sybase15, and copy the installation tar file to the unix machine, and uncompress and untar it. If there is no room any other place on the machine just place it into a subdirectory under sybase15. 2.       Make sure that the os has the necessary patches, if not arrange for them to be installed. 3.       You will need to ask the Unix sa to create a new raw device for you,  sybsystemdbdev of about 128MB for the new sybsystemdb database, e.g /dev/vg_fps/rlv_systemdbdev and also check the size of the sybsystemprocs device, it should be at least 160MB in size, if it’s less ask the Unix sa to create a new raw device for you, e.g /dev/vg_fps/rlv_sysprocsdev2 4.       Run setup to install ASE15 into its own separate directory (setup -console ), e.g sybase15, answer no to the question of licenses, we will fix this later on. So specify: –          Change the installation directory to e.g /progs/fps/sybase15 –          Select Full installation –          If asked about licenses select Continue installation without a license key –          Enterprise Edition –          SR :  Server License –          The answer yes to continue without license. –          No to configure email alerts Check that all the options are correct, especially the install directory, and click 1 to install the files. –          Answer NO to the remember password feature question. –          Deselect all the servers from configuration, we are not building the server now. 5.       You should install the latest patch for this version of ASE which can be downloaded from sybase.com, just remove the original installation files and replace them with the patch files then run the patch. Do the patching after the sybase server is upgraded. 6.       Make sure the existing Sybase server is sourced and is displayed in DSQUERY etc then run the script /progs/fps/sybase15/ASE-15_0/upgrade/preupgrade specify the sa password for the existing server to check that everything is in place for an upgrade. This programme will let you know if you need to increase any parameters etc and also if you need to expand any databases. Answer NO at the end when it asks if you want to proceed with the upgrade, this is just a precheck. You will most probably need to extend sybsystemprocs and sybsystemdb onto the new unix devices created earlier and also extend model to at least 4MB, in fact all user databases must be at least 4MB. If everything is fien then the last two lines should say: Preupgrade of Adaptive Server to 15.0 is complete. Upgrade eligibility test succeeded. 7.       You now need to add the following 3 environment variables to the SYBASE.csh file: OLDSYBASE= current SYBASE variable OLDSYBASE_ASE=current SYBASE_ASE OLDSYBASE_OCS=current SYBASE_OCS 8.       Source the new SYBASE.csh and create a resource file in $SYBASE/$SYBASE_ASE/init/sample_resource_files from the template where you fill in the values for the old $SYBASE path,  Sybase server name, Login (sa), and the password for current sa login on existing server. 9.       Turn off auditing on the existing sybase server and completely uninstall the sybsecurity database and sybaudit and reboot the Sybase server. Do an sp_configure “auditing”, 1 to make sure the auditing subsystem is switched off, VERY IMPORTANT!! 10.   You then run the upgrade with $SYBASE/$SYBASE_ASE/bin/sqlupgraderes –r $SYBASE/$SYBASE_ASE/init/sample_resource_files/<resource file>

Upgrading Rep server to REP15.0.2

Upgrade the RSSD database, start with the ID server.

  • Install the new software for replication server version 15 on your system. Put it in a new directory e.g /progs/torigodb/syb_rs15. After gunzipping and untarring the rep server tar ball run ./setup –console:
–          Accept the license agreement and select Full installation. –          Answer no to “Will licenses be obtained from the License server”. –          No to configure email alerts –          No to start a sample rep server
  • You can then install any required Sybase patches, e.g ESD1
  • Generate a license key (through sybase.subscribenet.com) and store it in the SYSAM-2_0/licences directory.
  • Copy any custom scripts over from the old repserver, e.g .cshrc, repscripts etc, and modify as necessary, i.e change SYBASE path etc.
  • Run “rs_helpsub” from the rssd database to verify that all subscriptions are valid. The status only needs to be Valid at the replicate replication server.
  • Copy the interfaces file (or interfaces) from the pre version 15 to the version 15 installation.
  • Log on the RSSD ASE server and grant the primary user the sa_role “grant role sa_role to <username>” if it doesn’t already have it. You can determine the name of the primary user by looking into the configuration file, at the keyword RSSD_primary_user.
  • Log on to the replication server and run “admin health” to see if everything is ok.
  • Shutdown the replication server, backup the RSSD database then restart the replication server.
  • Leave repserver running.
  • Set the environment variable $SYBASE to the new REP15 installation directory
  • Set the environment variable SYBASE_REP to REP-15_0
  • Start $SYBASE/$SYBASE_REP/install/rs_init (rs_init)
  • Select “Configure a Server product” -> “Replication server” -> “Upgrade an existing replication server”.
  • Answer all the questions. The configuration file should be the path and filename of the old config file.
  • Do Ctrl-A to start.

After a succesfull upgrade

  • Repoint the replication server paths to the new installation and copy the old runserver file to it’s new location. Make sure to change the paths in the runserver file to pick up the new version 15 executable and correct config files etc. Also copy over the old config file.
  • Fix or copy any other files bespoke files, You need to update the file $SYBASE/config/objectid.dat file and add a line similar to

Installing Heterogeneous Replication

To perform any heterogeneous replication, replication to and from two different datasources e.g Sybase to Oracle or DB2 to Oracle, you need to use Direct Connect or as it’s now being called Enterprise Connect Data Access. The below example is how to set up a connection to a new Oracle database but it can be applied to other databases as well obviously with some different scripts etc. In the below example Direct connect is already installed but unloading the software is not so difficult anyway, please refer to product manuals for details. 1.       First you will need to create a new Direct Connect process by creating a new config file in the $SYBASE/$SYBASE_DCO/install directory on the server where Direct Connect is installed. Just copy the existing .cfg file and RUN file and alter to them to reflect the new settings. So in the RUN file just change the servername and in the .cfg file add the new servername, username (in this case the oracle username), the path to the new errorlog file and the connect_string. 2.       Add an entry for this new Direct Connect process in the local interfaces file like: DCOCCDWP master tcp ether p-***-tm16 **** query tcp ether p-***-tm16 **** 3.       Also add this into the interfaces file for the REPServer and the ASE server which is acting as the RSSD for the Repserver. 4.       Add the Oracle connection information into the tnsnames.ora file which is in $SYBASE/$SYBASE_DCO/network/admin : DWMU = (DESCRIPTION = (ADDRESS_LIST = (ADDRESS = (PROTOCOL = TCP)(HOST = ***.**.***.**)(PORT = ****)) ) (CONNECT_DATA = (SERVICE_NAME = DWMU) ) ) 5.       You can now start the new Direct Connect process by running e.g startserver –f RUN_DCODWMU. 6.       You can check that it is working by trying to log into the oracle tablespace with e.g isql –Udefacto –SDCOCCDWP –P<oracle password>. 7.       You now need to go to the Repserver and make a copy of the following files in the $SYBASE/$SYBASE_REP/scripts directory : hds_oracle_udds.sql hds_clt_ase_to_oracle.sql hds_oracle_funcstrings.sql hds_oracle_setup_for_replicate.sql hds_oracle_connection_sample.sql For the first three scripts you need to edit the scripts and enter your RSSD database name and execute the scripts against the relevant ASE server housing the RSSD database for the Repserver, e.g isql –Usa – Sdhp_SOLP –DREP_dhpsolo_RSSD -i$SYBASE/$SYBASE_REP/scripts/DCOCCDWP_ hds_oracle_udds.sql The script hds_oracle_setup_for_replicate.sql needs to be copied over to the direct connect unix box and is run via Direct Connect against the replicate database, e.g isql -Ucm –P<oracle password> -SDCODWMU -ihds_oracle_setup_for_replicate.sql 8.       The last step is to create the connection to the Oracle database from the repserver and that is done by modifying the hds_oracle_connection_sample.sql script so that it looks similar to: create connection to DCOCCDWP.defacto set error class to rs_sqlserver_error_class set function string class to rs_oracle_function_class set username to ******* set password to ********** set ‘batch’ to ‘off’ go You then run this script against the repserver. 9.       To test that everything is ok you can try to connect to the oracle database form the replication server with e.g  isql –Udefacto –SDCOCCDWP –P<oracle password> If this connects then everything should work, at least in terms of connectivity. If it doesn’t work then you will probably have to open up the firewall both ways between the replication server port and the direct server port, e.g between pkg_solp (***.**.***.***) repserver port (2040) and dhp-tm16 (**.***.*.**) dco port (****). 10.   You then continue to define the rep defs, subscriptions etc.

How to add/update a table for replication

How to add/update a table for replication

–          First you need to create two scripts one to drop the existing subscription and replication definition and one to create the new table, replication definition and subscription. –          You also need to extract the exiting replicant table and also a user, e.g.  origomw, if requested and save this info. –          Before stopping replication send an email to operations to let them know of the planned downtime. –          At this point you should suspend the replication server connection from DB2, if it’s not already down and wait 5 minutes to make sure everything is applied. –          Do a select count(*) against an existing replicant table and compare this with a wc –l of the bcp file to make sure they match in number. For a new table this is not relevant. –          Bcp out the existing table just in case something goes wrong. –          Drop a user if that was requested –          Drop subscription, rep def in rep server: o   drop subscription DB2_ASN2_RESOURCE_NB_s2 for DB2_ASN2_RESOURCE_NB_d2 o   with replicate at origo_test.deFaktoReplica o   without purge o   go o o   drop replication definition “DB2_ASN2_RESOURCE_NB_d2″ o   go –          Drop the table in ASE o   Use database deFaktoReplica o   Go o   Drop table ASN2_RESOURCE_NB o   Go –          Create the table again –          Bcp the data in with a command like: o   bcp deFaktoReplica.dbo.ASN2_RESOURCE_NB  in /home/origo/deFaktoReplica/init/data/SYS3.ASN2.DDX -r ” \n” -t “\t” -c -Jiso_1 -Sorigo_test –Usa –P<password> -e bcplog.txt -z -b 10000 -m 100000 o   If you get problems check that the number of columns etc match and also check to see whether the row/tab delimiter is correct, sometimes there is an extra tab in the row delimiter, in that case try using –r “\t \n”. –          Add the primary key, triggers and other indexes etc into the table, and possibly also a rep server specific column like changed_date. –          Add the origomw user to the database. –          Add the replication definition and subscription. o   create replication definition “DB2_ASN2_RESOURCE_NB_d2” o   with primary at TSTA.P825RAD1 o   with primary table named “DB2_ASN2_RESOURCE_NB” o   with replicate table named “ASN2_RESOURCE_NB” o   ( o   “RESR_ELMT_ID”          int, o   “RECORD_EFF_END_DT”     datetime, o   “RECORD_EFF_END_TM”     datetime, o   “RECORD_EFF_STR_DT”     datetime, o   “RECORD_EFF_STR_TM”     datetime, o   “SRVC_LOC_ID”           numeric, o   “RESR_TYPE”             char(6), o   “RESR_ID”               char(22), o   “ACCT_ID”               numeric, o   “RESR_EFF_STR_DT”       datetime, o   “RESR_EFF_END_DT”       datetime, o   “ORD_ITEM_ID”           numeric, o   “FST_USG_DT”            datetime, o   “REINSTATE_CD”          char(4), o   “REINSTALL_DT”          datetime, o   “DISCN_CD”              char(4), o   “ASN_RESR_DISCN_DT”     datetime, o   “VERBAL_TRANSLATION”    char(20), o   “COMMENT_ID”            numeric, o   “REF_SEQ_NUM”           numeric, o   “INIT_INSTALL_DT”       datetime, o   “LAST_CHANGE_DT”        datetime, o   “RESR_GRP_TYP”          char(6), o   “ORD_ITEM_SEQ”          smallint, o   “PRIORITY_CD”           char(1), o   “SUB_STATUS_CD”         char(2), o   “PRMRY_COMP_CD”         char(6), o   “SECNDRY_COMP_CD”       char(6), o   “NUFS_NET_SRVC_TYP”     char(3), o   “NUFS_NUM_CAT”          char(2), o   “USER_ID”               char(8), o   “PREV_PHONE_NUM”        char(8), o   “RSU”                   char(6), o   “PRMRY_GRP”             smallint, o   “FSL”                   char(6), o   “CALL_INTERCEPT”        char(1), o   “RESR_SUB_GRP”          char(1), o   “OWNER_ACCT_ID”         numeric, o   “OWNER_SRVC_LOC_ID”     numeric, o   “PAYER_REF”             char(30), o   “PAYER_KURT_ID”         int, o   “CREATORS_REFERENCE”    char(25), o   “UPDT_LAST_MOD_TS”      datetime o   ) o   primary key o   ( o   “RESR_ELMT_ID”, o   “RECORD_EFF_END_DT”, o   “RECORD_EFF_END_TM” o   ) o   searchable columns (RECORD_EFF_END_DT) o   go o o o   define subscription DB2_ASN2_RESOURCE_NB_s2 o   for DB2_ASN2_RESOURCE_NB_d2 o   with replicate at origo_test.deFaktoReplica o   where RECORD_EFF_END_DT = ’31 dec 9999′ o   go –          Activate the subscription with: o   activate subscription DB2_ASN2_RESOURCE_NB_s2 o   for DB2_ASN2_RESOURCE_NB_d2 o   with replicate at origo_test.deFaktoReplica o   go –          Validate subscription with: o   validate subscription DB2_ASN2_RESOURCE_NB_s2 o   for DB2_ASN2_RESOURCE_NB_d2 o   with replicate at origo_test.deFaktoReplica o   go –          Resume replication and ask DB2 admin to start the Repagent on their side.

Installing RepServer 15

Installing RepServer 15

–          Untar the repserver installation files into a seperate directory and install the binaries etc by running setup –console. Answer no to email alerts and don’t enter license information. –          If you are migrating from an earlier version of Repserver and you have already migrated the ASE server then you need to make sure that the rep server logins are not expired, (e.g REP_dhtsolo_RSSD_prim and REP_dhtsolo_RSSD_maint). If they are expired, just try to login to the ase with them, you will need to change the passwords twice, first to a temp password and then back to the original. –          If you are using Norwegian language then you will also need to add the nocase_iso_1_nor.srt file into $SYBASE/charsets/iso_1 and tweak the $SYBASE/config/objectid.dat file by adding the following or similar line at the bottom of the collate section: o           1.3.6.1.4.1.897.4.9.3.148 = nocase_iso_1_nor Find the final number 148 by doing an sp_helpsort in the ase server and looking for the number associated with nocase_iso_1_nor –          Follow the regular steps for installing a repserver, i.e rs_init which should be fairly straight forward. When configuring the new rep server set the sort order to nocase_iso_1_nor

Installing a Sybase ASE15 Server

The following points apply for a new installation or an upgrade to an existing server. With an upgrade just create the new filesystem side by side the existing installation and migrate over the logins, config settings, user databases etc. In an upgrade the ideal is if you have enough space to run the 2 servers in parallel but if not then at least create the bare bones Sybase server (system databases) and then drop a user database on the existing sybase server, along with devices, and recreate them for the new server. It is important that you create environment variable for the new server which do not include any reference to the old server so you will need to hack the .cshrc and SYBASE.csh a little bit so that when you source the new .cshrc it loads the environment variables from scratch. Refer to the FPS Sybase server .cshrc file for examples. Also it is important that you give the new server a different port number from the old server. 1)      You first need to make sure that the Unix machine is ready and has all the required patches installed etc. 2)      The next step is to get the filesystem (5GB) and various raw devices created by Unix. You will need as an example at least the following raw devices in place before installing the Sybase server; /dev/vg_kapaks/rlv_master  (Mbytes) 224 /dev/vg_kapaks/rlv_sybsystemdev  (Mbytes) 128 /dev/vg_kapaks/rlv_sysprocsdev  (Mbytes) 320 /dev/vg_kapaks/rlv_temp01  (Mbytes) 2016 If you are doing an upgrade and space is limited then you could make the tempdb device smaller and then create a larger one later on. 3)      Next you need to install the actual server and for this you take the cd/dvd/tarball from Sybase and place it into $SYBASE/software. $SYBASE being the directory where you have decided to install Sybase, which must be different form any existing SYBASE installation. Unzip and extraxt the files. 4)      After you have untarred it you first set the SYBASe variable to where you want to install the server, e.g setenv SYBASE /progs/kapaks/sybase_15 and then run the setup by first starting the X server on your pc and then typing setup. You can also run it without a gui by typing setup –console. During the install you will be asked various questions including information about the license, just continue without license for now, we will put in the license later as we get a 30 day trial anyway. Choose Full installation, Enterprise edition, Server license and when asked say continue installation without license. Also answer no to configure email alerts, this can be done later if needed. Once all the software has been unloaded answer NO to remember ASE plugin password then you will be asked if you want to build the servers, go through the various fields filling in the relevant information and click on build. I would suggest initially building just the dataserver, backupserver and XP server. Other servers can be built later if required. Choose to custom configure all the servers. If this is a migration from an existing server remember to use the same port numbers as before and also to name the servers the same. Use the following answers, –          Answer Mixed to application type. –          2k to page size, VERY IMPORTANNT if this is a migration otherwise maybe choose 4k               for a completely new sybase server installation. –          Answer NO to enable PCI question. –          NO to optimise ASE configuration. –          Other values are fairly obvious hopefully. 5)      Hopefully this all goes fine and you now have a working Sybase server. You need to check whether the Sybase server needs to be localised to the Norwegian language, it almost certainly has to be. If you are migrating a server then you can easily check this in the old servers log file by looking for the line “SQL Server’s default sort order is”, if it says nocase_iso_1_nor then you need to install it for the new server. If it does then the next step is now to localise it to the Norwegian language and this is done by the following steps, if not you can jump over this step: –          First source the SYBASE.csh file by doing source ~/SYBASE.csh –          Copy the $SYBASE/charsets/iso_1/nocase_iso_1_nor.srt file over from another existing server and place it in the same folder on the new server i.e into  $SYBASE/charsets/iso_1/ folder and make sure it is readable by all. –          Start $SYBASE/ASE-15_0/bin/sqlloc (having first started exceeed on your pc) –          Select your Sybase server and enter the sa password, and then change the sort order to “Dictionary order, case insensitive, modified for Norwegian” and if necessary the default character set to iso_1 and click OK a few times. –          If the sqlloc application fails you can use sqllocres –r <resource file> (sqlloc.rs). The resource file can be found in $SYBASE/ASE-15_0/init/sample_resource_files. Just change the template resource file by putting in the Sybase server name, sa password and putting in sqlsrv.default_characterset: iso_1 and sqlsrv.sort_order: nocase_iso_1_nor 6)      The above step should be enough but some servers have an extra language module installed, for no apparently good reason. If you are migrating from an old server then you can check for this in the master..syslanguages table. If there is a row called Norwegian then you can install this on the new server by running: sp_addlanguage norwegian, norsk,                                                                                                                         ‘January,February,March,April,May,June,July,August,September,October, November, December’, ‘Jan,Feb,Mar,Apr,May,Jun,Jul,Aug,Sep,Oct,Nov,Dec’, ‘Monday,Tuesday,Wednesday,Thursday,Friday,Saturday,Sunday’, mdy,  7 7)      The next step is to extend the temporary database onto it’s own device 1GB is a good starting point, and you could also consider creating a few extra temporary databases. This is not necessary if you are installing Sybase ASE15.0.3 which asks for a special temporary device at install. You should at the least consider creating a special temporary database to be used by logins with the sa_role. 8)      You should probably put in a sa login password at this point for added security, you do this by logging into the server and typing sp_password NULL, <new password>. After this is done place the password into the $SYBASE/$SYBASE_ASE/install/$DSQUERY file. 9)      You now need to adjust various server parameters like memory, number of open objects etc, take these from an existing Sybase server and adjust as needed. The easiest way to do this is to look at the existing servers .cfg file and do something like a cat DHT_KAPAKS.cfg | grep -v DEFAULT to find the non default values. Ignore the sections for monitoring and security related, these will be set when installed later on. Also ignore the bit about buffer pools, you create this after the Sybase server is rebooted with the new parameters so do this now, just set the buffer pools to the same values on the server being migrated from. Also set the parameter number of open partitions to the sum of number of open objects and number of open indexes. 10)   At this point you can migrate over logins if you are doing a migration from an existing server. This will wary a bit but for example if you are upgrading from Sybase version 12 to 15 then you would do the following steps, for other ASE versions you might need to change the temp table a bit: –Create the temporary syslogins table on the new server sp_configure ‘allow updates to system tables’,1 go USE tempdb go –drop table dbo.temp_syslogins –go CREATE TABLE dbo.temp_syslogins ( suid        int      NOT NULL, status      smallint      NOT NULL, accdate     datetime      NOT NULL, totcpu      int           NOT NULL, totio       int           NOT NULL, spacelimit  int           NOT NULL, timelimit   int           NOT NULL, resultlimit int           NOT NULL, dbname      varchar(30)   NULL, name        varchar(30)   NOT NULL, password    varbinary(30) NULL, language    varchar(30)   NULL, pwdate      datetime      NULL, audflags    int           NULL, fullname    varchar(30)   NULL, srvname     varchar(30)   NULL, logincount  smallint      NULL, procid           int                NULL ) LOCK ALLPAGES Go n  You may need to tweak the column types depending on which version of Sybase you are importing from, check ther source sysloginroles table to double check. — Now bcp in the logins which was bcp’d out on the old Sybase server — bcp tempdb..temp_syslogins in tore_syslogins.out -Usa -P -SMICOS2 -n — Alter the table to add the new columns alter table tempdb..temp_syslogins add lastlogindate datetime NULL add crdate datetime NULL add locksuid int NULL add lockreason int NULL add lockdate datetime NULL go — Delete the sa and probe logins from the the temp_syslogins table use tempdb go delete from tempdb..temp_syslogins where name in (“sa”, “probe”) go — Delete existing logins which match by name delete from tempdb..temp_syslogins where name in ( select t.name from tempdb..temp_syslogins t, master..syslogins s where t.name =s.name ) go — Increase the suid’s. ONLY necessary if you are merging multiple sybase servers into one. update tempdb..temp_syslogins set suid = suid + 6150 go — Now copy the syslogins over to the master..syslogins table insert into master..syslogins select * from tempdb..temp_syslogins go — Create the sysloginroles table USE tempdb go CREATE TABLE dbo.temp_sysloginroles ( suid   smallint NOT NULL, srid   smallint NOT NULL, status smallint NOT NULL ) LOCK DATAROWS WITH EXP_ROW_SIZE=1 ON system Go n  You may need to tweak the column types depending on which version of Sybase you are importing from, check ther source sysloginroles table to double check. — BCP in the data — bcp tempdb..temp_sysloginroles in c1p16_sysloginroles.out -Usa -P -SMICOS2 -n — remove the roles for sa and probe logins delete from tempdb..temp_sysloginroles where suid <=2 go — Alter the table to make it compatible with ASE15 alter table tempdb..temp_sysloginroles modify suid int not NULL modify srid int not NULL go — Increase the suid’s ONLY necessary if you are merging multiple sybase servers into one. update tempdb..temp_sysloginroles set suid = suid + 6150 go — Copy the roles into master.sysloginroles insert into master..sysloginroles select * from tempdb..temp_sysloginroles go — The next steps relate to synchronising the suids after you have loaded the old              –database into the new server . –First on the original server check whether there are any aliases set up or whether        –there are any users who have a different name to their syslogin name with the             –following on the original Sybase database. select * from sysalternates go select l.suid,u.suid,l.name, u.name from master..syslogins l, sysusers u where l.suid = u.suid and l.name != u.name go –The following will resync the suids in a user database with the suid’s in syslogins. update sysusers set u.suid = l.suid from sysusers u, master..syslogins l where l.name=u.name 11)   You now need to set up the environment variables in .profile and .cshrc and also create the servername file containing the sa password, if you haven’t already done so, in the $SYBASE/ASE-15_0/install/ directory e.g for a Sybase server called MICOS1 this file would be called MICOS1 and would only contain the sa password, this is needed for the scripts to work. You can copy the environment variables from an existing Sybase server installation and just modify the values to your needs. 12)   You should now take a look at the $SYBASE installation folder and make sure that the interfaces, .cshrc , SYBASE.sh and SYBASE.csh files are all world readable. 13)   If this is a brand new machine then the next step is to generate the required licenses (always choose Un-served License). If this is just one server on an unclustered server, for example a test server, then that is fairly easy and the license file can be generated, (from https://sybase.subscribenet.com/control/sybs/login?nextURL=%2Fcontrol%2Fsybs%2Findex. If it is a clustered server then you need to create a multimode license file for each machine in the cluster and place it underneath the SYSAM folder. You create a multimode license file by answering “2” to the question of “Number of machines to license”. You then enter both the machines it can run on and generate and download the license file. The last step, for clustered and non-clustered Sybase servers is to place the licenses file under $SYBASE/$SYBASE_SYSAM2/licenses and if you want update the value for LM_LICENSE_FILE in the .profile and/or .cshrc file to point to it although this should not be necessary. 14)   Next we need to install the various standards that you migh thave, first you need to create the scripts directory (sc) etc, so just copy this from an existing server installation and place it in an appropriate place, best thing is to use a tar ball for this and just get rid of the unnecessary files afterwards like logs etc. 15)   Next we create the additional Sybase devices from the raw devices on Unix. 16)   The next step is to install several specific stored procedures etc: –          First create a new database called syb_drift with 20MB data and 5MB log CREATE DATABASE syb_drift ON data01=’20M’ LOG ON logg01=’5M’ go USE master go EXEC sp_dboption ‘syb_drift’,’trunc log on chkpt’,true go –          Type: cdhs cd sc inst (choose option 1 to install sybdrift, enter the sa password but answer no to                        “Vil du                 installere crontab” and “Vil du teste konsistens-sjekk, backup og                              overvaaking?”). –          Next you need to install the relevant sp_thresholdaction stored procedure and this depends on whether or not you are using SQL Backtrack to backup the databases. If you are using SQL Backtrack then choose the following stored proc: USE sybsystemprocs go IF OBJECT_ID(‘dbo.sp_thresholdaction’) IS NOT NULL BEGIN DROP PROCEDURE dbo.sp_thresholdaction IF OBJECT_ID(‘dbo.sp_thresholdaction’) IS NOT NULL PRINT ‘<<< FAILED DROPPING PROCEDURE dbo.sp_thresholdaction >>>’ ELSE PRINT ‘<<< DROPPED PROCEDURE dbo.sp_thresholdaction >>>’ END go USE sybsystemprocs go IF OBJECT_ID(‘dbo.sp_thresholdaction_old’) IS NOT NULL BEGIN DROP PROCEDURE dbo.sp_thresholdaction_old IF OBJECT_ID(‘dbo.sp_thresholdaction_old’) IS NOT NULL PRINT ‘<<< FAILED DROPPING PROCEDURE dbo.sp_thresholdaction_old >>>’ ELSE PRINT ‘<<< DROPPED PROCEDURE dbo.sp_thresholdaction_old >>>’ END go create procedure sp_thresholdaction_old @dbname         varchar(30), @segmentname    varchar(30), @space_left     int, @status         int as declare @devname varchar(100), @before_size int, @after_size int, @before_time datetime, @after_time datetime, @error int if @segmentname != (select name from syssegments where segment = 2) begin print “THRESHOLD WARNING: database ‘%1!’, segment ‘%2!’ at ‘%3!’ pages”, @dbname, @segmentname, @space_left end go EXEC sp_procxmode ‘dbo.sp_thresholdaction_old’, ‘unchained’ go IF OBJECT_ID(‘dbo.sp_thresholdaction_old’) IS NOT NULL PRINT ‘<<< CREATED PROCEDURE dbo.sp_thresholdaction_old >>>’ ELSE PRINT ‘<<< FAILED CREATING PROCEDURE dbo.sp_thresholdaction_old >>>’ go create procedure sp_thresholdaction @dbname         varchar(30), @segmentname    varchar(30), @space_left     int, @status         int as declare @devname varchar(100), @before_size int, @after_size int, @before_time datetime, @after_time datetime, @error int, @cmd1 varchar(5000), @cmd2 varchar(5000) set @cmd1 = ‘$DT_SBACKTRACK_HOME/bin/dtsbackup ${DTPHYSICAL}/’ + @dbname + ‘ -log_only | tee $BACKUPKAT/ch/backtrack_logg’ set @cmd2 = ‘$BACKUPKAT/sc/m_sback_log ‘ + @dbname –if @segmentname != (select name from syssegments where segment = 2) begin print “THRESHOLD WARNING: database ‘%1!’, segment ‘%2!’ at ‘%3!’ pages”, @dbname, @segmentname, @space_left exec xp_cmdshell @cmd1 exec xp_cmdshell @cmd2 exec xp_cmdshell @cmd1 end go EXEC sp_procxmode ‘dbo.sp_thresholdaction’, ‘unchained’ go IF OBJECT_ID(‘dbo.sp_thresholdaction’) IS NOT NULL PRINT ‘<<< CREATED PROCEDURE dbo.sp_thresholdaction >>>’ ELSE PRINT ‘<<< FAILED CREATING PROCEDURE dbo.sp_thresholdaction >>>’ go If however you are not using SQL Backtrack install the following version of sp_thresholdaction: USE sybsystemprocs go IF OBJECT_ID(‘dbo.sp_thresholdaction’) IS NOT NULL BEGIN DROP PROCEDURE dbo.sp_thresholdaction IF OBJECT_ID(‘dbo.sp_thresholdaction’) IS NOT NULL PRINT ‘<<< FAILED DROPPING PROCEDURE dbo.sp_thresholdaction >>>’ ELSE PRINT ‘<<< DROPPED PROCEDURE dbo.sp_thresholdaction >>>’ END go USE sybsystemprocs go IF OBJECT_ID(‘dbo.sp_thresholdaction_old’) IS NOT NULL BEGIN DROP PROCEDURE dbo.sp_thresholdaction_old IF OBJECT_ID(‘dbo.sp_thresholdaction_old’) IS NOT NULL PRINT ‘<<< FAILED DROPPING PROCEDURE dbo.sp_thresholdaction_old >>>’ ELSE PRINT ‘<<< DROPPED PROCEDURE dbo.sp_thresholdaction_old >>>’ END go create procedure sp_thresholdaction_old @dbname         varchar(30), @segmentname    varchar(30), @space_left     int, @status         int as declare @devname varchar(100), @before_size int, @after_size int, @before_time datetime, @after_time datetime, @error int if @segmentname != (select name from syssegments where segment = 2) begin print “THRESHOLD WARNING: database ‘%1!’, segment ‘%2!’ at ‘%3!’ pages”, @dbname, @segmentname, @space_left end go EXEC sp_procxmode ‘dbo.sp_thresholdaction_old’, ‘unchained’ go IF OBJECT_ID(‘dbo.sp_thresholdaction_old’) IS NOT NULL PRINT ‘<<< CREATED PROCEDURE dbo.sp_thresholdaction_old >>>’ ELSE PRINT ‘<<< FAILED CREATING PROCEDURE dbo.sp_thresholdaction_old >>>’ go create procedure sp_thresholdaction @dbname         varchar(30), @segmentname    varchar(30), @space_left     int, @status         int as declare @devname varchar(100), @before_size int, @after_size int, @before_time datetime, @after_time datetime, @error int, @cmd1 varchar(255) select @cmd1 = ‘$BACKUPKAT/sc/dump_t ‘ + @dbname –if @segmentname != (select name from syssegments where segment = 2) begin print “THRESHOLD WARNING: database ‘%1!’, segment ‘%2!’ at ‘%3!’ pages”, @dbname, @segmentname, @space_left exec xp_cmdshell @cmd1 end go EXEC sp_procxmode ‘dbo.sp_thresholdaction’, ‘unchained’ go IF OBJECT_ID(‘dbo.sp_thresholdaction’) IS NOT NULL PRINT ‘<<< CREATED PROCEDURE dbo.sp_thresholdaction >>>’ ELSE PRINT ‘<<< FAILED CREATING PROCEDURE dbo.sp_thresholdaction >>>’ go 17)   The next step is to configure the new mda montables and that is done as follows: sp_configure “enable monitoring”,1 go sp_configure “SQL batch capture”,1 go sp_configure “max SQL text monitored”,100000 go sp_configure “sql text pipe active”,1 go sp_configure “sql text pipe max messages”,10000 go sp_configure “object lockwait timing”,1 go sp_configure “per object statistics active”,1 go sp_configure “statement cache size”, 10000 go sp_configure “enable stmt cache monitoring”,1 go sp_configure “deadlock pipe max messages”, 10000 go sp_configure “deadlock pipe active”, 1 go sp_configure “errorlog pipe active”,1 go sp_configure “errorlog pipe max messages”,10000 go sp_configure “wait event timing”,1 go sp_configure “statement statistics active”,1 go sp_configure “process wait events”, 1 go sp_configure “plan text pipe active”, 1 go sp_configure “plan text pipe max messages”, 10000 go sp_configure “statement pipe max messages”, 10000 go sp_configure “statement pipe active”,1 go 18)   Install any extra stored procedures, these can be found from an existing Sybase server. E.g sp__mda_hot_tables, sp_mda_io etc. 19)   At this stage you can create (for load) and start loading the user databases, using sql backtrack remote load procedure, for example from existing sybase server run; dtsrecover /home/solo1/datatools/sbackups.physical/dhp_SOLP/commissiondb -server TORIGODB -database commissiondb –user sa –password <pass> -copyover Remember to perform  the suid resync section above. If you haven’t done so previously  you need to first create all the required sybase data devices from the raw partitions which should already be on the Unix machine. 20)   Install dbccdb –          Run sp_plan_dbccdb to find the recommended size of the dbccdb and create it with for example: USE master go CREATE DATABASE dbccdb ON data11=’1500M’ LOG ON log01=’500M’ go USE master go EXEC sp_dboption ‘dbccdb’,’trunc log on chkpt’,true go USE dbccdb go CHECKPOINT Go –          Run the scripts/installdbccdb script in Unix using isql. –          Run the following sql with the value being the max number for processes displayed by sp_plan_dbccdb: sp_configure “number of worker processes”, 6 go –          If you haven’t already done so earlier create a 150MB 16k memory pool in the cache used by dbccdb, usually just default data cache. EXEC sp_poolconfig ‘default data cache’,’150M’,’16K’ go EXEC sp_poolconfig ‘default data cache’,’16K’,’wash=30M’ Go –          Create the workspaces from the maximum recommended values as follows: Use dbccdb Go sp_dbcc_createws dbccdb, ‘default’, scan_dbccdb, scan , ‘750M’ go sp_dbcc_createws dbccdb, ‘default’, text_dbccdb, text , ‘200M’ go –          Configure dbccdb for each of the user databases as follows, just use the max values for all the db’s: sp_dbcc_updateconfig tlfbank, ‘max worker processes’, ’12’ go sp_dbcc_updateconfig tlfbank, ‘dbcc named cache’, ‘default data cache’,                                          ‘200M’ go sp_dbcc_updateconfig tlfbank, ‘scan workspace’, scan_dbccdb go sp_dbcc_updateconfig tlfbank, ‘text workspace’, text_dbccdb go 21)   The next thing to install is auditing and unique sa logins and the details for this can be found in the document Auditing and Unique Logins. 22)   It is a good idea to create a second tempdb to reduce the load on the tempdb and that is done as follows: create temporary database tempdb2 on tempdb_dev2=’1024M’ log on tempdb_log=’524M’ go sp_tempdb ‘add’, ‘tempdb2’, ‘default’ go sp_tempdb ‘show’ go 23)   You will also need to install SQL Backtrack if that is required. 24)   When you come to load the actual databases at migration or upgrade time make sure that you run update index stats against all the user tables AND all the system tables except syslogs and sysgams otherwise SQL Backtrack will run very slowly. 25)   The final step is to set up all the cron jobs to run, here just look at what is currently configured on the existing or similar servers. In some cases (check with Unix sa) you need to create a copy of the crontab in the $HOME directory for clustering etc. You do this with a command similar to; crontab -l > $HOME/crontab.sybgunda where sybgunda is the username. 26) Another thing you may find is that SQL Backtrack stops working if you have changed the ip address of the machine, this might be because all the TSM (OBSI settings) have changed. To fix this you need to find out what the new settings are in the dsm.sys file and then probably create a new adsm dump pool.

How to drop a database when drop database fails

How to Drop a Database When drop database Fails

Follow the steps in this section to drop a database when drop database fails. Do not use these steps unless directed to do so by this book, or unless there is no critical data in the database.

1.Log in as the “sa”.

2.Check to make sure the database has been marked “suspect.” The following query produces a list of all databases which are marked suspect:

1> select name from master..sysdatabases

2> where status &256 = 256

3> go

3.If the database is marked “suspect”, go to step 4. If it is not marked “suspect”, mark it in one of the following ways:

a.Execute the sp_marksuspect stored procedure discussed under “How to Mark a Database “suspect””, and restart Adaptive Server to initialize the change.

b.Use the procedure below:

1> sp_configure”allow updates”, 1

2> go

1> use master

2> go

1> begin transaction

2> update sysdatabasesset status = 256

3> where name = “database_name”

4> go

Verify that only one row was affected and commit the transaction:

1> commit transaction

2> go

Reset the allow updates option of sp_configure:

1> sp_configure “allow updates”, 0

2>go

Restart Adaptive Server to initialize the change.

4.Remove the database:

1> dbccdbrepair(database_name,dropdb)

2> go

dbcc dbrepair sometimes displays an error message even though it successfully drops the database. If an error message occurs, verify that the database is gone by executing the use database_name command. This command should fail with a  911 error, since you dropped the database. If you find any other error, contact Sybase Technical Support.

How to move the master database to a new device

This error occurs when you try to extend the master database onto a device other than the master device. It is recommended that you keep user objects out of the master database. If you keep user databases off the master device, you allow space in case the master database needs to grow. In addition, if you ever need to rebuild the master device, it will be easier if it does not contain user databases. Adaptive Server users can move any “home-grown” system procedures that start with “sp_” to sybsystemprocs (by dropping them from the master database and creating them in sybsystemprocs). Extend the master database only if absolutely necessary! If you are sure you must increase the master database size and have no room on the current master device, use the following procedure to remove user databases from the master device. Move User Databases
  • Dump the user databases with the dump database command.
  • Rename the dumped databases on the master device with sp_renamedb.
  • Re-create the databases with their original names on another device with create database. Be sure they are created exactly like the old databases, to avoid 2558 and other errors. Refer to Error 2558 for more information.
  • Load the dumps with load database.
  • Use the online database command for each database to make the databases available for use.
  • Check the databases in their new location to make sure the load was successful (that is, perform a simple query with isql), and if everything loaded successfully, drop the old databases from the master device.
You can now try to increase the size of the master database on the master device with the alter database command. Increase Master Device Size If the master device contains only the master database and the master device is too small, then use the following procedure: Warning! Altering the master device is extremely risky! Avoid it if at all possible. Be familiar with the recovery methods in “System Database Recovery” in case you lose your master database or master device.
  • Back up the master database with the dump database command.
  • Save the contents of key system tables such as sysdatabases, sysdevices, sysusages, and syslogins.  Make a note of these values.  Also make a note of the path to the dump device in sysdevices.
  • Use the buildmaster utility to build a new master device with enough extra space so that you will never need to increase the master device again. When buildmaster completes, a new master database will exist on the new master device. The buildmaster executable is found in bin, so use ./buildmaster and follow the prompts.
  • You now need to create a new runserver file which points to this new master device (the -d option). And start up the server with this new runserver file.
  • Expand the size of the new master database with the alter database command, if necessary, so that it matches the size of the dumped master database(get this info from the original sysusages table where the size is in 2k blocks, the alter database command uses sizes in MB).
  • Execute the following command in isql:
1> select name, high from master..sysdevices 2> where name = “master” 3> go and note the “high” value for the master device. Shutdown the server.
  • Add the –m option to the runserver file to start Adaptive Server in single-user mode.
  • Allow updates to the system catalog:
1> sp_configure “allow updates”, 1 2> go
  • Change the value for srvnetname in sysservers from SYB_BACKUP to the name of your backup server.
  • Load the dump of the master database, using load database master from <full path name>.
  • Reset the “high” value in master..sysdevices:
1> begin transaction 2> go 1> update master..sysdevices 2> set high = <value of high from step 5> 3> where name = “master” 4> go
  • If the previous update affected only one row, commit the transaction.
  • Restart Adaptive Server.
  • Turn off allow updates:
1> sp_configure, “allow updates”, 0 2> go
  • Edit the new runserver file to take it out of single user mode, i.e remove the –m option and restart the server, if this all works fine (leave it for a while) then you can remove the original master device and its related run server file.

How to perform a load froma remote backup server

How to perform a load from a remote backup server A step by step guide: In this guide there are assumed to be two servers TROPHY_1103 and TRIDENT_1103. 1.                  Create a backup server for TROPHY_1103 called TROPHY_1103_BACK 2.                  Create a backup server for TRIDENT_1103 called TRIDENT_1103_BACK (For info on creating backup servers refer to reference manuals) 3.                  Log into TROPHY_1103 and execute the following two commands:                          sp_addserver SYB_BACKUP, TROPHY_1103_BACK sp_addserver TRIDENT_1103_BACK (If the SYB_BACKUP part doesn’t work do a sp_dropserver SYB_BACKUP     first) 4.                  Log into TRIDENT_1103 and execute the following two commands: sp_addserver SYB_BACKUP, TRIDENT_1103_BACK sp_addserver TROPHY_1103_BACK 5.                  Take a look at the interfaces file for TROPHY_1103 and make a note of the entry for TROPHY_1103_BACK.  This info needs to be entered into the Runserver file for TRIDENT_1103 except the line which starts master.  The entry for TRIDENT_1103_BACK in the TRIDENT_1103 Runserver file needs to be entered into the TROPHY_1103 Runserver file again taking out the master line, leaving the query line. 6.                  Next make sure that both servers TROPHY_1103 and TRIDENT_1103 are set up for remote procedure calls by checking that sp_configure “allow remote access” has a run value set to 1, if not issue the following command sp_configure “allow remote access”, 1 7.                  Test that the two servers can communicate by performing the following procedure calls,                                                                                       from TROPHY_1103; TRIDENT_1103_BACK…sp_who And from TRIDENT_1103; TROPHY_1103_BACK…sp_who 8.         The two servers are now set up to allow remote backups, this can be issued             from either server (in this example I’m  performing it from TROPHY_1103)    with the following command: Load database from “<give the full pathname to the tape or disk device used         by TRIDENT_1103 ASE>” at TRIDENT_1103_BACK.