On 7/26/2010 2:05 PM, Keith Lofstrom wrote: >> Jason Boxman wrote: >>> Been using Dirvish since like 2003 and I've never seen this behavior before. > > On Mon, Jul 26, 2010 at 10:40:58AM +0100, Dave Howorth wrote: >> You could also up the verbosity wherever possible and add extra print >> statements (pre- post- options, perhaps) and maybe run the various jobs > > I second Dave's suggestion about running client and server pre and post > scripts (I do all four). I use these scripts to gather extra information, > most importantly the output of "df" and "sfdisk -l" so I can rebuild > disks more easily, and do forensics. That is also where I cancel out > that night's dirvish run to a client if an rsync is already running, > and build a line of status information to save in a terse backup log. > > In the pre-client, I also run a silly little C program I wrote called > "bigalloc", which builds a huge memory array and then releases it. > This pushes most seldom-used programs out of ram and into swap, > making rsync run faster. Or at least it did 3 years ago; perhaps > the kernel and rsync have improved so this is not necessary. In any > case, if the program fails before rsync runs, I learn something useful. > > Here is a very old writeup that describes some of my scripts: > http://www.keithl.com/linuxbackup.html > Needs updating and correcting - six years old! - but it should give > you some ideas. BTW, JW disliked the illustration, which is why it > is not on the website.
What I ultimately did was simply change the run-time back to 3 a.m. Clearly there must be some kind of strange bug if changing the time fixes everything. _______________________________________________ Dirvish mailing list [email protected] http://www.dirvish.org/mailman/listinfo/dirvish
