Ralf Gross wrote: > Ralf Gross schrieb: >> Les Mikesell schrieb: >>> On 5/26/2010 3:41 PM, Ralf Gross wrote: >>>> Ralf Gross schrieb: >>>>> write(1, "N\2\0\7\5\3lvs\r\0\0\0\r\0\0\0lvmiopversion8\5"..., 594) = 594 >>>>> select(1, [0], [], NULL, {60, 0}) = 0 (Timeout) >>>>> select(1, [0], [], NULL, {60, 0}) = 0 (Timeout) >>>>> select(1, [0], [], NULL, {60, 0}) = 0 (Timeout) >>>>> select(1, [0], [], NULL, {60, 0}) = 0 (Timeout) >>>>> select(1, [0], [], NULL, {60, 0} >>>>> >>>>> >>>>> smells like a time out, but I don't know where. I found a couple of >>>>> messages >>>>> with similar output in the list archives, but none of them had a solution >>>>> yet. >>>> *grr* >>>> >>>> I only traced the Xfer PID, not the PID. BackupPC_dump seems to be >>>> active and comparing the file list with the pool and I see high cpu >>>> load. >>>> >>>> I'm sure that I haven't seen that as I abortet the backup before. >>>> Now I'll have will wait until tomorrow morning... >>> Until the 2nd full completes, the server side has to uncompress the >>> stored copy to compute the checkums on existing files. And there may be >>> some quirk about switching from tar to rsync that I've forgotten. Maybe >>> the 1st run will add the checksum cache for files you already have. >> >> The full rsync is still running sind 5/26 21:00. I'll report back when >> it's done. > > > Ok, the first rsync full backup (488) completed. It took 500min. longer than > the last tar full backup (482). > > Backup Type Filled Level Start Date Duration/mins Age/days > 482 full yes 0 5/19 02:05 3223.2 11.5 > 483 incr no 1 5/21 07:49 89.6 9.2 > 484 incr no 2 5/22 03:05 136.4 8.4 > 485 incr no 3 5/23 03:05 119.1 7.4 > 486 incr no 4 5/24 03:05 111.4 6.4 > 487 incr no 1 5/25 03:05 165.9 5.4 > 488 full yes 0 5/26 21:00 3744.2 3.7 > 489 incr no 1 5/29 12:15 394.1 1.1 > 490 incr no 2 5/30 03:05 190.8 0.4 > > I'm not sure if the checksum caching will compensate this in after the 3rd > backup. Anything else I could do to tune rsync? >
You could force a full to start on Friday evening so weekly scheduling will keep the full runs on weekends if they take more than a night to complete. Depending on how much daily change you have, you might want to set incremental levels for the intermediate runs. A more extreme change would be to edit Rsync.pm to not add the --ignore-times option on fulls. I haven't needed this myself yet but I think it would make a big difference in speed - at the expense of not checking files for unlikely but possible differences. -- Les Mikesell lesmikes...@gmail.com ------------------------------------------------------------------------------ _______________________________________________ BackupPC-users mailing list BackupPC-users@lists.sourceforge.net List: https://lists.sourceforge.net/lists/listinfo/backuppc-users Wiki: http://backuppc.wiki.sourceforge.net Project: http://backuppc.sourceforge.net/