Thanks for your fast response. Really seems to be the network that is so slow. Looks like somebody is heavily using it at the moment. But this still doesn't explain why the storage daemon won't wait for the data to be transmitted and then begin writing to the tape. I'm quite sure I activated spooling. I inserted "Spool Data = yes" only in the Job ressource (not in bacula-sd because there is no such option according to the manual, bacula-sd also confirms that by refusing to start if I inserted this directive in the Storage ressource). After starting the according job i get the message: 15-Aug 12:52 debian-sd: Spooling data ... But still the daemon writes immediately to the tape. The spool file exists and grows slowly (slow because of the connection). When it's full (i.e. has the size of the data to be backuped) the sd states it's despooling but this takes one or two seconds because the data has been written to the tape all the time.
I'm using Debian and so I use version 1.38.11. Do you think this is an issue? The database also shouldn't be a bottleneck. I backuped 5000 small files from the machine connected to the tape drive within seconds with no performance loss (50-60 MB/s). I think the issue is just the spooling... > > Now i backup some files over the network without spooling and get about > 300 > kb/s. If I activate spooling I get 700 kb/s. (Yesterday I got 4 MB/s but this > is > still ridiculously low...) > Is this a full backup job? Do you have a gigabit network? What version > of bacula are you using? Have you properly indexed your database? If > you do not have a gigabit network is your database on the same server > as the director and storage? Are you backing up thousands of small > files? Do you have software compression on? Are your computers less > than 3 years old or all > 2GHz. > > > Isn't the storage daemon supposed to wait until all the data is on the > hard drive and then start writing to the tape? > Yes. Bacula 2.1.X waits till the entire spool file is filled before it > writes to the tape. There has been work to allow the tape drive to > work while the client is running (to improve single backup > performance) but I do not think that made it into the current bacula > which is version 2.2.0. > > >This doesn't to be the case at all. A spool file is created but also > the writing to the tape starts at the same time so writing speed still > is very low. > > Are you sure that spooling is on? I believe you have to activate it in > the job and the bacula-sd. It will show that it is running if you do a > status storage while the job is running. How big is your spool file? > > My spool file is to 2GB for my nearly 100% gigabit network with my > LTO2 autochanger and with spooling I get between 4MB/s to 25MB/s which > is dependent on a lot of factors. Incremental and diferential jobs are > usually much slower than Full because the client spends a lot of time > finding the files to backup. If I backup a job with > 1,000,000 small > files I generally get a lower much rate than backing up a few files > that total the same size as the 1,000,000 small files. > > I am running my director, database and storage on three different > machines with the director and storage being dual processor Opteron > 248 boxes with 4GB of memory and the database (postgres) is on an > athlon64 3200 with 2GB of memory. All are running 64 bit gentoo with > most servers running bacula 2.1.X and clients (windows and linux) > running versions of bacula from 1.38 to 2.0.X. > > John > > John ___________________________________ NOCC, http://nocc.sourceforge.net ------------------------------------------------------------------------- This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now >> http://get.splunk.com/ _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
