On 6/22/06, Kern Sibbald <[EMAIL PROTECTED]> wrote: > Hello, > > I am not likely to accept the Feature request as it stands (perhaps after some > changes) for the following two reasons: > > 1. You mention that this feature will prolong the tape's lifetime, but I see > no impovement in this regard over the current code. While it is despooling and additional data would be spooled in the background, the writing would be longer and probably the tape would not have to be stopped and restarted so many times. In case if the fd can read as fast as the sd despools or if the speeds are close enough, there would also be a possibility that it would write the whole tape without stopping. Or imagine, in case of incremental cases, well the tape drive usually is a lot slower than the hard drive. So while it is searching for the next changed file, it (the sd) is despooling data, when the fd finds new files, it sends it to the sd ... if the files are big enough, the sd would not need to stop the tape at all. > > 2. Once the spool file fills, it is full. To then restart filling it while it > is despooling seems to me to be rather impossible on the one hand, and would > lead to a lot of race conditions on the other hand. First, despooling > requires reading the spool file and writing it to tape. *Only* when the full > spool file has been read can it be truncated and thus free up space for > additional spooling. Second, if one were somehow to figure out out to > release blocks at the beginning of a file, you would then be in a condition > where you had a lot of excess activity on the computer where the despooling > would be constantanly blocking and deblocking the new FD writes. This would > be *extremely* inefficient. > > One possibility in continuing spooling would be to introduce "multiple" spool > directories or some way to partition a spool directory so that it would say > fill half-way and then begin despooling while the other half could continue > spooling. What about a configurable setting? Splitting the spool files into blocks of the specified size (example 1GB). When the sd reads one of the spool files (and writes it ofcourse) it deletes it and reads the next one if it is available (and in worst case scenario, if the fd hasnt finished sending data and the specified block size hasnt been reached the file has .part extension or something like that? after it fills the block or fd has already finished its task, it is renamed into a file expected by the sd). > > On Wednesday 21 June 2006 18:31, Rumko wrote: > > Item n: Spooling data while despooling it > > Origin: Rumko <[EMAIL PROTECTED]> > > Date: 21 June 2006 > > Status: > > > > What: Currently when fd runs out of data or maximum spool size is > > reached, the sd starts despooling. While it is despooling, the fd does > > not spool the data until all data has been despooled. For performance > > reasons (and it would prolong tape's lifetime), the fd should spool > > data while the data is being despooled by the sd. > > > > Why: Better performance on tape drives (writing would be > > prolonged, because while writing the data, more would be spooled and > > made available to the sd) and longer lifetimes for the tapes. > > > > Notes: I was thinking of 2 solutions. Perhaps a FIFO? But there > > comes a question, (since i dont know much about fifos) how much data > > can be put into a fifo at one time and if theres a limit, what happens > > when you reach it? Will the fd still be able to start piping data into > > the fifo after it has already filled it once? ... The second solution > > would be separate files. Perhaps the sd should inform the fd 5 seconds > > (which should be settable in the config files) before running out of > > data that the fd should start spooling data into a new file and leave > > the old one exclusively to the sd (so that there would be no conflicts > > or any other problem that would fall into this realm). > > > > > > _______________________________________________ > > Bacula-users mailing list > > Bacula-users@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/bacula-users > > -- > Best regards, > > Kern > > ("> > /\ > V_V >
Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642 _______________________________________________ Bacula-users mailing list Bacula-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/bacula-users
