Paul Bijnens wrote:
> Gene Heskett wrote:
>
>> On Friday 04 October 2002 11:15, Joshua Baker-LePain wrote:
>>
>>> On Fri, 4 Oct 2002 at 10:54am, c white wrote
>>>
>>>
>>>> i've been running amanda for over a year now and every time i
>>>> get a report it tells me that the esitmated time for the job is
>>>> 3 minutes, when it actually take almost 1.5 hours, why is this?
>>>
>>>
>>> That's "Estimate" time, not estimated time. "Estimate time" is
>>> how long it takes amanda to run a faux backup on all your
>>> filesystems to get an estimate of how big the various level
>>> backups would be, and choose what level to do on every
>>> filesystem. Look in /tmp/amanda/sendsize*debug on your clients
>>> to see what amanda is doing during the estimate phase.
>>
>>
>>
>> That must be a fairly small system this gentleman is estimating.
>> Here, with 37 entries in the disklist, and serialized by use of a
>> spindle number, its about an hour, then another 1.5 to 2.5 to finish,
>> some of which is time in the compressor, so the drive sits idle for a
>> few minutes occasionally. Dropping the spindle number useage so all
>> the dumpers can run at the same time only seems to hammer the hard
>> drive harder without significantly reducing runtime. Just my middle
>> of the night observation though.
>>
>
>
> When using dump instead of GNUtar you could get in the minute-range
> instead of in the hour-range. (But dump has other problems, so
> sometimes you need to use tar instead. Like me.)
>
i'm using tar and its doing 4gb nightly with almost 20 different disk
entries (i started the whole thread by asking the quick question)
