Idees
------
+Despool attributes later
+ - use spool_attribute parameter (0,1,2)
+ foreach spool_file (spool_files)
+ ask to director ok/discard
+ despool_attr
+ end_of_despool
+
+Continious Backup Linux
+ - stap
+http://www.redhat.com/magazine/011sep05/features/systemtap/
+http://sourceware.org/systemtap/
+
+Deduplication:
+ proba de collision en fonction de la taille du hash
+ http://en.wikipedia.org/wiki/Birthday_attack
+
Cryptage :
- Pouvoir configurer les algos de cryptage
- Pouvoir choisir de crypter dans le fileset
Bconsole :
p Ajouter l'historique dans la bconsole
- Ajouter une commande pour bloquer un drive
+ - Ajouter de la completion automatique sur les commandes
Sauvegarde SAN :
- Avoir un SD sur chaque FD
- générer un format simple UNIXSTAMP|label|unit|value
- possibilité d'ajouter ça dans la base de donnée ou dans rrd
+Exchange backup:
+http://www.petri.co.il/brick_level_backup_of_mailboxes_by_using_exmerge.htm
+
API :
- apr (apache portable runtime) pour les exemples de chargement
----
bweb :
+ - Support sqlite
+ SELECT strftime('%Y-%m-%d', Job.StartTime) FROM Job LIMIT 1;
+ x Voir les groupes d'un client
+ - Overview, pouvoir choisir entre la job_old et l'autre
+ - utiliser des noms de table plus proche (brestore -> bweb ?)
- Balloon
Btw., the vertical axis was the number of files
while the size of the ballon is the volume (in MB
- Ajouter la liste des fichiers (quand il n'y en a pas trop)
- Ajouter un mode qui compte le nombre de fichier sous bfileview
+ x Ajouter une estimation de progression du backup basé sur le nombre
+ de fichier et le nombre de Mo
x Ajouter un bouton suivant/precedant dans la vue des logs
x Ajouter la liste des medias qui vont/ont expirer et les pruner
x Fixer les purge/prune de multiple volumes
p modifier l'organisation de la table version (pour pouvoir ajouter
les versions de bweb par exemple)
- utiliser la meme fonction pour read_close_session et append_close_session
- - pb dans le message de chargement d'une bande pendant une resto
+ x pb dans le message de chargement d'une bande pendant une resto
Please mount Volume "R40175" or label a new one for
Pas le bon message pour une resto (label a new one)
jobs normaux. Le patch précédent ne changeait pas le MaxConcurrentJobs
comme il faut à la fin du backup.
- - Accurate backup
+ x Accurate backup
o Envoyer la liste de tous les fichiers au client dans un format simple
/path/ LSTAT # un / a la fin pour un repertoire
/path/file LSTAT
p Ajouter une option pour avoir la table de stat
Enable Statistic = yes/no
Statistic retention = 4 years
-
-
- o Utiliser une alist dans les runscripts
+ o use mmap to map hash ? (on 32b, we are limited to 1 or 2GB)
+
+#ifndef _WIN32
+ ef->data = mmap(NULL, ef->data_size, PROT_READ,
+ MAP_SHARED, fileno(ef->fp), 0);
+#else
+ fm = CreateFileMapping((HANDLE) _get_osfhandle (fileno(ef->fp)),
+ NULL,
+ PAGE_READONLY,
+ 0,
+ 0,
+ NULL);
+ ef->data = MapViewOfFile(fm,
+ FILE_MAP_READ,
+ 0,
+ 0,
+ ef->data_size);
+ CloseHandle(fm);
+#endif
+
+ ef = eet_internal_read(ef);
+ if (!ef)
+@@ -892,11 +862,7 @@
+ free(ef->header);
+ }
+
+#ifndef _WIN32
+ if (ef->data) munmap((void*)ef->data, ef->data_size);
+#else
+ if (ef->data) UnmapViewOfFile (ef->data);
+#endif
+
+if (ef->fp) fclose(ef->fp);
+
+ - Accurate backup (kern)
+1. Run bconsole
+2. Dir -> FD run job
+*3. FD does a normal backup and at the same time makes a list of all files on
+the system (in the FileSet), marking which ones were just now backed up.
+4. For each file backed up send attributes/data to SD. Note, this is done
+during step 3 above. Minor difference, the connection with the SD is not
+dropped at the end of the backup -- see later.
+*5. Send the list of all files including those backed up to the Dir
+ --> Send to SD and DIR at the same time ?
+ filed/backup.c/encode_and_send_attributes
+
+6. Dir computes files and deleted files.
+7. Dir sends list of additional files (new files) to backup, and list of files
+deleted.
+8. FD does backup of files Dir says to save.
+9. FD sends SD attrs of backed up files
+10. FD sends SD delete records for files Dir wants deleted.
+*11. FD connection to SD can be closed, this part of the backup is done.
+*12. FD sends new list of files just backed up to Dir
+*13. Dir adds newly backed up files to previous list sent by FD
+*14. Dir "batch" inserts complete new list in the catalog (I forgot the name
+of the new table). Note this table has nothing to do with the File table.
+*15. Dir deletes previous list in catalog.
+*16. Dir does normal batch insert of attributes from SD, but must handle
+deleted records. Note, this will probably happen at the same time as the
+items 13-15.
+
+
+
+ - TODO:
+ 0001088: volume FirstWritten attribute is set to time of mount request, not time of first write
+ Description When a Bacula job requests mounting a tape volume that is not present in the drive,
+ once the tape is mounted, its FirstWritten attribute is set to the time when the volume was requested.
+ Consequently, if the job has been waiting longer than the maximum use duration of the volume,
+ the volume is promoted to Used immediately because the maximum use duration has apparently expired before
+ the use has even started.
+
+To avoid that, the FirstWritten attribute should be set to the time the volume was mounted (= the current time when the setting takes place).
+
+ x Backup a file that is not in accurate list (change NOCHG to LINK, FILE, etc..)
+ * Manage JobFiles (Deleted compte pour 1 ?)
+
+ x Utiliser le check_accurate dans find_one et declencher le save_file
+ si besoin en desactivant le incremental.
+ x ne va pas marcher avec le strip path (la recherche est faite avant le strip path)
+ * on peut utiliser le champs LStat de la base pour noter que le fichier est supprimé...
+
+ CREATE TEMPORARY TABLE btemp2 AS (
+ SELECT max(FileId) as FileId, PathId, FilenameId
+ FROM (SELECT FileId, PathId, FilenameId FROM File WHERE JobId IN (39867,40341)) AS F
+ GROUP BY PathId, FilenameId )
+
+ SELECT Path.Path, Filename.Name, File.FileIndex, File.JobId, File.LStat
+ FROM (
+ SELECT max(FileId) as FileId, PathId, FilenameId
+ FROM (SELECT FileId, PathId, FilenameId FROM File WHERE JobId IN (11,13)) AS F
+ GROUP BY PathId, FilenameId
+ ) AS Temp
+ JOIN Filename ON (Filename.FilenameId = Temp.FilenameId)
+ JOIN Path ON (Path.PathId = Temp.PathId)
+ JOIN File ON (File.FileId = Temp.FileId)
+ WHERE FileIndex > 0
+
+
+ SELECT File.FileIndex, Path.Path, Filename.Name, File.LStat
+ FROM btemp2 JOIN Path USING (PathId) JOIN Filename USING (FilenameId)
+ JOIN File USING (FileId)
+ WHERE File.FileIndex > 0
+
+ DROP TABLE btemp2
+*/
+/*
+SELECT DISTINCT ON (PathId, FilenameId) FileIndex, Path, Name, LStat
+ FROM File JOIN Filename USING (FilenameId) JOIN Path USING (PathId) WHERE JobId IN (40341)
+ ORDER BY PathId, FilenameId, JobId DESC
+*/
+
+ - .api mode:
+Some ideas :
+- Every dates have to be in ISO format
+YYYY-MM-DD HH:MM:SS
+- JobLevel, JobStatus, etc.. have to use C constant T,R,a,A,f...
+- Sizes are always in bytes (no suffix)
+- Numbers have to be used without commas
+- If we change (add) something, we must always add new elements
+at the end.
+
+For director status on running jobs, it will be great to display :
+JobId, Client name, Job Name, Level, Start Time and Status
+
+
+ x Utiliser une alist dans les runscripts
RunScript {
console = "xxxx"
console = "aaaa"
}
- x fix segfault when config files are empty
o cleanup bextract to use filed code
+
projects / bacula / bacula / blobdiff