Both interfaces accomplish the same three steps. The steps are to select jobs to restore from, give the user the opportunity to select files/directories, then indicate details such as the host and path to restore to and trigger the job to run.
Start the standard restore procedure by pressing the restore button in the task bar. There are also two options in the joblist context sensitive menu to start a restore. They are Restore From Time or Restore From Job.
This restore method is intended as a high performance option. It is a server side process. These interfaces assist the user in utilizing the text based restore capabilities of the standard console. It interprets the text to display the information in a way that simplifies the restore procedure.
The Opening interface allows the user to choose selection criterion to inform the server how to determine the set of backup job ids to use in the restore. This best possible set is he most recent full backup, the most recent differential backup done since the most recent full, and all the incremental backups done since the most recent full or differential. Then the server uses this set of jobs to create a file structure that is the most recent version of each file found in this job list.
The second interface allows the user to browse this file structure and choose the files and directories to restore. This is done in an explorer type interface with a directory tree on the left. In the right pane of a splitter is a table showing a list of files and directories that are the contents of the directory selected in the left pane. The user can mark and unmark either with the buttons on the top or by double clicking on the check mark icon to toggle whether an item is selected or not. Double clicking an item which is a directory on a part of the table which is not the check icon opens that directory. Clicking 'OK' completes the selection process.
The third step is the Restore Run interface. It's purpose is to allow the user to inform the bacula server details of the host and path to restore to, the replacement rules, when to restore and what priority to give the restore job.
This restore interface is NOT intended to perform major restores of directory structures with large numbers of directories and files. It should work, however it is a chatty interface. This is due to the number of sql queries made of the server which is proportional to the number of files and directories selected plus the number of exceptions to defaults selected. It IS intended to allow the user to browse for specific files and choose between the different versions of those files that are available in the catalog to restore.
The interface contains a horizontal splitter. The bottom pane contains some controls for the interface. The top portion contains a vertical splitter with 4 panes for viewing the cataloged information. The left pane is for viewing and further sub selecting of jobs. The second pane is for viewing the directory tree. The third is for viewing a list of files in a directory that has been selected. Then lastly the fourth pane is for viewing a table of versions of a single file that has been selected from the file table.
The version browser accomplishes the three restore steps differently.
To select jobs and populate the directory tree, press the refresh button. The job table contains selected jobs. The selection criterion of the three dropdowns and the two limits are used as the filtering criterion for populating the job table the first time the refresh button is pushed. After the refresh button has been pushed, the job table has check marks that can selects and unselects jobs. Re-pressing the refresh button does one of two things. What occurs is dependent on if the controls in the bottom pane display the same data as the previous time the refresh button was pressed. If changed the jobs table is repopulated from the selection criterion. If unchanged any jobs that have been unchecked are excluded from the process of selecting directories, files and versions. The directory tree does get repopulated when the refresh button is pushed. There is a text label underneath the refresh button to inform the user as to which occurs when refresh is pressed.
The user can browse the directory tree and click on a directory folder which then populates the file table with the files that are contained in the selected directory path. Selecting or unselecting a directory does also select or unselect all files and all directories in the tree hierarchy beneath it. If there are any exceptions already selected beneath that directory, those exceptions do get deleted.
With the file table populated, the user can unselect a file in a selected directory and also select a file in an unselected directory.
With a file selected the version table populates with all the instances a file has been written to tape. The user can choose a specific version of a file for restore and override the default which is to restore the most recent version.
Pressing the restore button initiates a procedure preparing to perform the restore of the requested files. The same Restore Run interface that was the third step in the standard restore is then displayed. It allows the user to instruct the bacula server of the details of what host and what path to restore the files to. This part of the restore does take control of the connection to the server and does not allow any other communication to the server by the other interfaces.
There are two progress bars that appear when refreshing or after pressing Restore. These indicate to the user the time it may take to complete any tasks that could take a long time period.
If you have used the version browser with a large database, you may have noticed that the performance can begin to be quite slow. A way to improve the response time of the database server is to add indexes that will assist a couple of the specific queries that are made.
For postgresql add 2 indexes with the following commands in psql:
CREATE INDEX file_filenameid_jobid ON file USING btree (filenameid,
jobid);
CREATE INDEX file_pathid_idx ON file USING btree (pathid);
For mysql add 2 indexes with the following commands in mysql:
CREATE INDEX file_filenameid_jobid ON File (FilenameId, JobId);
CREATE INDEX file_pathid_idx ON File (PathId);
There is one way to make the first of those two indexes perform just a
little better. It is to create a partial index. First, at least one backup
must be in the database that has at least one directory. Then in psql or mysql
perform the command:
SELECT FilenameId FROM Filename WHERE name='';
Use the results of this command and replace for XXX in the following command:
CREATE INDEX file_filenameid_jobid2 ON File (JobId) WHERE FilenameId=XXX;
This index will use less disk space and will perform better. Don't forget to
remove the index it replaces, file_filenameid_jobid.
If you have sqlite and would be willing to test out the creation of these indexes to see if they work, please let me know the commands.