Hvrlogrelease
Name
hvrlogrelease - Manage DBMS logging files when not needed by log-based capture.
Synopsis
hvrlogrelease [optfile] [-options]
Description
hvrlogrelease works differently depending on database class applied: it either manages log files (for Oracle and Ingres) or frees log file blocks (for SQL Server). Database class is given by one of -oracle_sid, -ingres_db or -sqlserver_db options, which must be provided.
Functionality for Oracle
hvrlogrelease can be used to manage Oracle archive files so that they are available if needed by HVR's log-based capture jobs. The default behavior is that hvrlogrelease monitors the 'primary' archive directory where the Oracle archiver process writes these files and makes a copy in a private directory whenever one appears. It then removes these copies when the HVR capture jobs will no longer need their contents.
An alternative behavior is activated by option -purge; in this case no copies are made and hvrlogrelease simply deletes Oracle archive file from the 'primary' archive directory once it sees these files will no longer be needed by HVR capture jobs. This is useful if archiving was only enabled for HVR's log-based capture and are not needed for backup and recovery.
It relies on 'log release' files that are re-created by log-based capture jobs after each replication cycle. These files are in $HVR_CONFIG/files and contain timestamps which allows hvrlogrelease to see which archive files could still be needed by HVR replication.
When hvrlogrelease is making private copies of archive files (option -purge not defined), they are placed in a special directory using a file system 'hard link' to avoid i/o overhead. The path is defined by action Capture /ArchiveLogPath. If that is not defined then a default path is derived by adding the suffix _hvr to the Oracle archive destination directory.
The files copied by hvrlogrelease keep their original 'base name', so parameter /ArchiveLogFormat cannot be used.
Command hvrlogrelease with option -purge can be used with action Capture /ArchiveLogPath /ArchiveLogOnly to remove files which the capture job no longer needs. In this case, an additional option -force must be used.
Functionality for Ingres
hvrlogrelease can be used to manage Ingres journal files so that they are available if needed by HVR log-based capture jobs. The default behavior is that hvrlogrelease monitors the directory where the DBMS creates these files, and makes a copy in a private directory whenever one appears. It then removes these copies when the HVR capture jobs will no longer need their contents.
An alternative behavior is activated by option -purge; in this case no copies are made and hvrlogrelease simply deletes DBMS logging files from the archive or journal directory once it sees these files will no longer be needed by HVR capture jobs. This is only useful if journaling was enabled for HVR's log-based capture and are not needed for backup and recovery.
It relies on 'log release' files that are re-created by log-based capture jobs after each replication cycle. These files are in $HVR_CONFIG/files and contain timestamps which allows hvrlogrelease to see which DBMS logging files could still be needed by HVR replication.
When hvrlogrelease makes private copies of Ingres journal files (option -purge not defined) they are placed in special directory using a file system 'hard link' to avoid i/o overhead. The path is defined by environment variable $HVR_LOG_RELEASE_DIR. If that is not defined then a default path is derived by adding the suffix _hvr to the database's journal directory.
Functionality for SQL Server
hvrlogrelease resets log file blocks from 'taken' into 'free' state. Such blocks can be later reused to prevent excessive log file growth. That log 'truncation' may be needed in the following situations:
- coexistence, when several brands of replication tools need to read same log file;
- multi-capture, when several capture jobs read the same log file;
- in order to reset the bunch of used log file blocks when log has not being truncated for a long period.<br/>
First two scenarios require configuring HVR capture job with automatic log truncation turned off. Then log file needs to be truncated by executing hvrlogrelease command from according job ran by schedule or any other means. For more information, see Managing SQL Server Log file truncation.
Notes
- hvrlogrelease must be scheduled to run under the DBMS owner's login (e.g. oracle or ingres) , whereas hvrmaint must run under the HVR's login.
- When hvrlogrelease is installed on an Oracle RAC, then the $HVR_CONFIG directory must be shared between all nodes. Directory $HVR_TMP (if configured) should not be shared. Command hvrlogrelease should then be scheduled to run on all nodes, but with 'interleaved' timings. For example 0, 20, 40 minutes after each hour on one node and 10, 30, 50 minutes after each hour on the other node.
- Command hvrlogrelease does not support archive files located inside Oracle ASM. In this situation the RMAN must be configured to retain the archive files for sufficient time for HVR.
Options
This section describes the options available for command hvrlogrelease.
| Parameter | Description |
|---|---|
| -email=addr1[;addr2] | Email output from hvrlogrelease output to addr1 [and addr2]. Requires either option -smtp_server or option -mailer. Multiple email addresses can be specified, either using a single -email option with values separated by a semicolon or using multiple -email options. |
| -email_only_errors | Only send an email if hvrlogrelease encounters an error. |
| -email_activity | Only send an email if hvrlogrelease encounters an error or purged or compressed any file. |
| -email_from=from | Specify a from address in email header. |
| -env=NAME=VALUE | Set environment variable, such as $HVR_HOME. This option can be repeated to set different variables. Values for $ORACLE_SID and $HVR_CONFIG should be defined with special options -oracle_sid and -hvr_config respectively. |
| -help | Provide options list with service information to create UI dialog. |
| -hvr_config=dir | Check for 'log release' files in this $HVR_CONFIG directory. This option must be supplied at least once. It can also be supplied several times if multiple HVR installations capture log changes from a single database. hvrlogrelease will then purge DBMS logging files only after they have been released by all the HVR installations. If value dir contains an asterisk (*) then all matching directories are searched. |
| -ingres_db=db | Only check 'log release' files for Ingres database db. If value db contains an asterisk (*) then all Ingres databases are matched. This option can be supplied several times for different databases. Note that hvrlogrelease extracts the path $II_SYSTEM from matching log release files, so it could affect journal files which are not in the current Ingres installation. |
| -logrelease_expire=Nunits | Instruct command hvrlogrelease to ignore any 'log release file' that are too old. Value units can be days, hours or minutes. For example, value 4days could be defined because if the capture job has not run for four days then the replication backlog is so great that a refresh will be needed and the DBMS logging files can be discarded. |
| -mailer=cmd | Mailer command to use for sending emails, instead of sending them via an SMTP server. Requires option -email. String %s contained in cmd is replaced by the email subject and string %a is replaced by the intended recipients of the email. The body of the email is piped to cmd as stdin. |
| -oracle_sid=sid | Only check for 'log release' files for Oracle instance sid. If value sid contains an asterisk (*) then all Oracle instances are matched. This option can be supplied several times for different instances. Note that hvrlogrelease extracts the value of $ORACLE_HOME from matching log release files, so it could affect archived redo files which are not in the current Oracle installation. |
| -output=fil | Append hvrlogrelease output to file fil. If this option is not supplied then output is sent to stdout. Output can also be sent to an operator using option -email. |
| -state_dir=dir | Create files hvrlogrelease.pid and hvrlogrelease.dirs in provided directory dir. If not supplied these files will be created in directory set in environment variable $HVR_CONFIG. |
| -purge | Purge (i.e. delete) old DBMS logging files (Ingres journals and Oracle archived redo logfiles) and backup files (e.g. Ingres checkpoints) once the 'log release' files indicate that they are no longer needed for any HVR replication. If the 'log release' file is absent or unchanged then nothing is purged, so operators must purge journals/archived redo files manually. This is only useful if journaling/archiving was only enabled for HVR's log-based capture and are not needed for backup and recovery. |
| -smtp_server=server | SMTP server to use when sending email. Value server can be either a node name or IP address. Requires option -email. |
| -smtp_user=user | Username user for authentication SMTP server if needed. |
| -smtp_pass=pass | Password pass used for authentication on the SMTP server if needed. |
| -user=login/pass | Database authentication login and password. |
| -sqlserver_db=db | Mark SQL Server log file blocks free for database db. |
| -verbose | Write extra logging to output file. |
Deleting Ingres Journals with CKPDB
Command hvrlogrelease can encounter problems if Ingres command ckpdb is used with option -d (delete journal files). This is because Ingres will create new journal files when 'draining' the log file and then delete them so quickly that hvrlogrelease does not have time to make a copy. The solution is to create a customized Ingres 'checkpoint template' file which calls hvrlogrelease from within ckpdb. Two lines of this file need to be changed: those labeled PSDD and PSTD. The following steps use a Perl expression to change them to call hvrlogrelease with specific option file (this file must also be created).
$ cd $II_SYSTEM/ingres/files $ perl -pe 's/(PS[DT]D:\s*)/$1hvrlogrelease \$HVR_CONFIG\/files\/hvrlogrelease.opt;/' cktmpl.def >cktmpl_hvr.def $ ingsetenv II_CKTMPL_FILE $II_SYSTEM/ingres/files/cktmpl_hvr.def
Examples
Example 1 - manage Oracle archive file copies
The following can be saved in option file /opt/hvrlogrelease.opt so that private copies of any Oracle archive files from instance ORA1020 are available when needed by HVR log-based capture:
-env=HVR_HOME=/opt/hvr410/hvr_home -hvr_config=/opt/hvr410/hvr_config -oracle_sid=ORA1020 -logrelease_expire=3days -email_from=hvr@prod.mycorp.com -email=bob@mycorp.com;jim@mycorp.com -email_only_errors
If Oracle command rman is also configured to remove old redo logfiles, then hvrlogrelease must be scheduled to run first so that it sees every file before that file is removed by rman. This can be done by scheduling both hvrlogrelease and rman in a single crontab line.
00,15,30,45 * * * * su oracle -c "/opt/hvr_home/bin/hvrlogrelease /opt/hvrlogrelease.opt >> /tmp/hvrlogrelease.log 2>&1 && rman @delete_archives >> /tmp/delete_archives.log 2>&1"
Example 2 - manage Ingres journal file copies
The following option file can be used to maintain private copies of any Ingres journal files for database mydb so that they are available when needed by HVR log-based capture:
-env=HVR_HOME=/opt/hvr410/hvr_home -hvr_config=/opt/hvr410/hvr_config -ingres_db=mydb -email_from=hvr@prod.mycorp.com -email=bob@mycorp.com;jim@mycorp.com -email_only_errors
Example 3 - mark SQL Server log blocks as free
The following option file can be used to move Log Sequence Number replication pointer for database mydb to mark virtual log file blocks as free:
-sqlserver_db=mydb -user=my_login/my_pwd -output=c:\hvr\hvr_config\log\hvrlogrelease-hub-mydb.out
Files
