flush-threshold-dumped not exactly working

[[email protected]]

We are very happy with 3.2.0 ... thanks for the great work.

but ...

with:

flush-threshold-dumped 25
flush-threshold-scheduled 50
taperflush 50
autoflush yes

[tape] length 24576 mbytes

After ~6500m have been dumped, the first filesystem (233 mbytes) of 21 filesystems (~30500 mbytes) is written to tape. After that nothing more is written to tape until dumping is complete.

Afterward, everything seems good. That's important!

Shouldn't the writing to tape continue after filesystem 1 is written?

I'll admit to incomplete understanding of these *flush* parameters.

[ghosts]

We also have the same problem.

We use the default values

flush-threshold-dumped 0
flush-threshold-scheduled 0
taperflush 0

For our understanding of this values, amanda should start with the flush after the first dump is finished and should flush all following dumps as soon as they are finished. Just like amanda did it in the past.

We can see this since we switched to amanda-3.2 and i think it has todo with the changes in server-src/driver.c and the calculations for my_flush_threshold_dumped, my_flush_threshold_scheduled and my_taperflush variables.
But after I'm no real programmer and there are debug settings here to make amanda to write out these value i can't really say this for sure ...


I also owe this forum a description of our setup ...
We use 2 amanda servers with each about 12 TB on iSCSI disk space for Virtual Disks ( chg-disk ) and 6 Holding disks, on which at least one whole backup run finds place.
The 2 Servers have to backup 300 clients by now

We have a simple,classical structure with 18 V-Tapes, so we have one tape for each backup run, after we have this much clients amanda can manage the usable disk space by tape pretty good with full and incremental runs

but with this bug, the whole runtime of amanda takes 2 more hours, because amanda don't flush out the already done backups until the last, most time really slow, client has finished it's run.

[martineau]

A fix was committed this morning

Can you try it?

[ghosts]

for us it seems good now
we made some test runs and all looked fine
thx

[FHogenhout]

I apologize for this probably silly question but where can I get the fix ?

I used the RedHat 5 rpm from the Amanda-download page for installation and this is from 18th October.

[ghosts]

I added the patch from the checkin for revision 3608
[url]http://amanda.svn.sourceforge.net/viewvc/amanda?view=revision&revision=3608[/url]