| |
Home |
![[Table of Contents]](/dwicons/b_toc.gif)
|
Reference Manual: Commands |
![[-]](/dwicons/i_colpse.gif)
| Chapter 1 Commands |
|
| dump transaction |
|
| Usage |
Table 1-24 describes the commands and system procedures used to back up databases and logs.
To do this | Use this command |
Make routine dumps of the entire database, including the transaction log. | dump database |
Make routine dumps of the transaction log, then truncate the inactive portion. | dump transaction |
Dump the transaction log after failure of a database device. | dump transaction with no_truncate |
Truncate the log without making a backup. Then copy the entire database. | dump transaction with truncate_only dump database |
Truncate the log after your usual method fails due to insufficient log space. Then copy the entire database. | dump transaction with no_log dump database |
Respond to the Backup Server's volume change messages. | sp_volchanged |
You cannot dump to the null device (on UNIX, /dev/null).
You cannot use the dump transaction command in a transaction.
When using 1/4-inch cartridge tape, you can dump only one database or transaction log per tape.
You cannot issue dump the transaction log while the trunc log on chkpt database option is enabled or after enabling select into/bulk copy/pllsort and making minimally logged changes to the database with select into, fast bulk copy operations, default unlogged writetext operations, or a parallel sort. Use dump database instead.
Warning!
Never modify the log table syslogs with a delete, update, or insert command.
If a database does not have a log segment on a separate device from data segments, you cannot use dump transaction to copy the log and truncate it.
If a user or threshold procedure issues a dump transaction command on a database where a dump database or another dump transaction is in progress, the second command sleeps until the first completes.
To restore a database, use load database to load the most recent database dump; then use load transaction to load each subsequent transaction log dump in the order in which it was made.
Each time you add or remove a cross-database constraint, or drop a table that contains a cross-database constraint, dump both of the affected databases.
Warning!
Loading earlier dumps of these databases can cause database corruption.
You cannot dump from an 11.x Adaptive Server to a 10.x Backup Server.
You cannot have Sybase dumps and non-Sybase data (for example, UNIX archives) on the same tape.
You cannot dump a transaction with no_log or with truncate_only if the database has offline pages.
After device failure, use dump transaction with no_truncate to copy the log without truncating it. You can use this option only if your log is on a separate segment and your master database is accessible.
The backup created by dump transaction with no_truncate is the most recent dump for your log. When restoring the database, load this dump last.
When a database does not have a log segment on a separate device from data segments, use dump transaction with truncate_only to remove committed transactions from the log without making a backup copy.
Warning!
dump transaction with truncate_only provides no means to recover your databases. Run dump database at the earliest opportunity to ensure recoverability.
Use with truncate_only on the master, model, and sybsystemprocs databases, which do not have log segments on a separate device from data segments.
You can also use this option on very small databases that store the transaction log and data on the same device.
Mission-critical user databases should have log segments on a separate device from data segments. Use the log on clause of create database to create a database with a separate log segment, or alter database and sp_logdevice to transfer the log to a separate device.
Use the with standby_access option to dump transaction logs for loading into a server that acts as a warm standby server for the database.
When you use with standby_access to dump the transaction log, the dump proceeds to the furthest point in the log at which all earlier transactions have completed and there are no records belonging to open transactions.
You must use dump tran[saction]...with standby_access in all situations where you will be loading two or more transaction logs in sequence and you want the database to be online between loads.
After loading a dump made with the with standby_access option, use the online database command with the for standby_access option to make the database accessible.
Warning!
If a transaction log contains open transactions and you dump it without the with standby_access option, version 11.9.2 does not allow you to load the log, bring the database online, then load a subsequent transaction dump. If you are going to load a series of transaction dumps, you can bring the database online only after a load that was originally dumped with standby_access or after loading the entire series.
Warning!
Use dump transaction with no_log only as a last resort, after your usual method of dumping the transaction log (dump transaction or dump transaction with truncate_only) fails because of insufficient log space. dump transaction with no_log provides no means to recover your databases. Run dump database at the earliest opportunity to ensure recoverability.
dump transaction...with no_log truncates the log without logging the dump transaction event. Because it copies no data, it requires only the name of the database.
Every use of dump transaction...with no_log is considered an error and is recorded in Adaptive Server's error log.
If you have created your databases with log segments on a separate device from data segments, written a last-chance threshold procedure that dumps your transaction log often enough, and allocated enough space to your log and database, you should not have to use this option. If you must use with no_log, increase the frequency of your dumps and the amount of log space.
Transaction log dumps are dynamic--they can take place while the database is active. They may slow the system slightly, so run dumps when the database is not being heavily updated.
Use dump database immediately after creating a database to make a copy of the entire database. You cannot run dump transaction on a new database until you have run dump database.
Develop a regular schedule for backing up user databases and their transaction logs.
dump transaction uses less storage space and takes less time than dump database. Typically, transaction log dumps are made more frequently than database dumps.
Use thresholds to automate backup procedures. To take advantage of Adaptive Server's last-chance threshold, create user databases with log segments on a separate device from data segments.
When space on the log segment falls below the last-chance threshold, Adaptive Server executes the last-chance threshold procedure. Including a dump transaction command in your last-chance threshold procedure helps protect you from running out of log space. For more information, see sp_thresholdaction.
You can use sp_addthreshold to add a second threshold to monitor log space. For more information about thresholds, see the System Administration Guide.
You can specify the dump device as a literal, a local variable, or a parameter to a stored procedure.
You can specify a local dump device as:
A logical device name from the sysdevices system table
An absolute path name
A relative path name
The Backup Server resolves relative path names using Adaptive Server's current working directory.
Dumping to multiple stripes is supported for tape and disk devices. Placing multiple dumps on a device is supported only for tape devices.
When dumping across the network, specify the absolute path name of the dump device. The path name must be valid on the machine on which the Backup Server is running. If the name includes any characters except letters, numbers, or the underscore (_), enclose it in quotes.
Ownership and permissions problems on the dump device may interfere with use of dump commands. sp_addumpdevice adds the device to the system tables, but does not guarantee that you can dump to that device or create a file as a dump device.
You can run more than one dump (or load) at the same time, as long as they use different dump devices.
If you issue a dump transaction command without the init qualifier and Backup Server cannot determine the device type, the dump transaction command fails. For more information, see the System Administration Guide.
You must have a Backup Server running on the same machine as your Adaptive Server. The Backup Server must be listed in the master..sysservers table. This entry is created during installation or upgrade and should not be deleted.
If your backup devices are located on another machine so that you dump across a network, you must also have a Backup Server installed on the remote machine.
Dumping a log with the init option overwrites any existing files on the tape or disk.
Dump file names identify which database was dumped and when the dump was made. If you do not specify a file name, Backup Server creates a default file name by concatenating the following:
Last seven characters of the database name
Two-digit year number
Three-digit day of the year (1- 366)
Hexadecimal-encoded time at which the dump file was created
For example, the file cations930590E100 contains a copy of the publications database made on the 59th day of 1993:
Figure 1-4: File naming convention for transaction log dumps
The Backup Server sends the dump file name to the location specified by the with notify clause. Before storing a backup tape, the operator should label it with the database name, file name, date, and other pertinent information. When loading a tape without an identifying label, use the with headeronly and with listonly options to determine the contents.
Dump volumes are labeled according to the ANSI tape-labeling standard. The label includes the logical volume number and the position of the device within the stripe set.
During loads, Backup Server uses the tape label to verify that volumes are mounted in the correct order. This allows you to load from a smaller number of devices than you used at dump time.
When dumping and loading across the network, you must specify the same number of stripe devices for each operation.
On UNIX systems - the Backup Server requests a volume change when the tape capacity has been reached. After mounting another volume, the operator notifies the Backup Server by executing the sp_volchanged system procedure on any Adaptive Server that can communicate with the Backup Server.
If the Backup Server detects a problem with the currently mounted volume (for example, if the wrong volume is mounted), it requests a volume change by sending messages to either the client or its operator console. The operator responds to these messages with the sp_volchanged system procedure.
By default (noinit), Backup Server writes successive dumps to the same tape volume, making efficient use of high-capacity tape media. Data is added following the last end-of-tape mark. New dumps can be appended only to the last volume of a multivolume dump. Before writing to the tape, Backup Server verifies that the first file has not yet expired. If the tape contains non-Sybase data, Backup Server rejects it to avoid destroying potentially valuable information.
Use the init option to reinitialize a volume. If you specify init, Backup Server overwrites any existing contents, even if the tape contains non-Sybase data, the first file has not yet expired, or the tape has ANSI access restrictions.
Figure 1-5 illustrates how to dump three transaction logs to a single volume. Use:
init to initialize the tape for the first dump
noinit (the default) to append subsequent dumps
unload to rewind and unload the tape after the last dump
Figure 1-5: Dumping three transaction logs to a single volume
At the beginning of a dump transaction, Adaptive Server passes the primary device name of each logical log device to the Backup Server. If the primary device has been unmirrored, Adaptive Server passes the name of the secondary device instead. If the named device fails before Backup Server completes its data transfer, Adaptive Server aborts the dump.
If you attempt to unmirror a named log device while a dump transaction is in progress, Adaptive Server displays a message. The user executing the disk unmirror command can abort the dump or defer the disk unmirror until after the dump completes.
dump transaction with truncate_only and dump transaction with no_log do not use the Backup Server. These commands are not affected when a log device is unmirrored, either by a device failure or by a disk unmirror command.
dump transaction copies only the log segment. It is not affected when a data-only device is unmirrored, either by a device failure or by a disk unmirror command.
|
|