On 2017-10-24 12:28, Stefan G. Weichinger wrote:
Am 2017-10-24 um 13:38 schrieb Austin S. Hemmelgarn:
On 2017-10-22 13:38, Stefan G. Weichinger wrote:
After or before I additionally can do something like:

amvault myconf --dest-storage --latest-fulls archive

correct?
I think so, but I'm not 100% certain.

oh ;-)

An additional hurdle is that the customer wants to use WORM tapes for
archive, so I should get that right at the first run to not waste any tapes

Perhaps create a temporary virtual tape library for testing that the archiving schedule works as expected?  This is what I generally do when testing changes at work (although I usually do it using a copy of the main configuration so that I don't confuse the planner for the production backups with half a dozen runs in one day).

Sure, that would be good, but I don't have that much disk space available.

I am currently trying to wrap my head around the tuning of these parameters (and understand the exact meaning by reading the man page):

flush-threshold-dumped
flush-threshold-scheduled
taperflush

I had lev0 of all DLEs in the holding disk and both flush-threshold values on 400 -> I thought this would keep data for 4 tapes inside the disk, but no, some lev0 backups were flushed to primary storage already.

Maybe I set up a VM with 2 vtape changers and play around there to learn and understand.
Based on what you're saying you want, I think you want the following in your config:

    flush-threshold-dumped 400
    flush-threshold-scheduled 400
    taperflush 400
    autoflush yes

The first two control flushing during a run, while taperflush controls flushing at the end of a run. To get the flushing to actually happen, you then need autoflush set to yes (and amanda will complain if it's not set to yes and taperflush is more than zero).

Now, I'm not 100% certain that will work, as I've not done this type of thing myself (at work, we just use the holding disk as a cache so that we can finish dumps as quickly as possible without our (slow, parity-raid backed) persistent storage being the bottleneck, and at home I don't use it since I don't need parallelization and I don't have any disks that are faster than any others), but based on what I understand from the documentation, I'm pretty sure this should do it.

Reply via email to