On Friday 20 June 2003 08:49, Jason Edgecombe wrote:
>Joshua Baker-LePain wrote:
>> On Tue, 17 Jun 2003 at 3:23pm, Jason Edgecombe wrote
>>
>>>  I'm trying to figure out why may backups and estimages take so
>>> long. I have a dual-processor machine with 4GB RAM. / and /boot
>>> are two software RAID1 drives and /home is a RAID5 hardware
>>> drive. My tape drive is an Overland Neo2000 LTO2 with 30 tape
>>> library.
>>>
>>>I am using a 200GB holding disk on an isolated drive dedicated to
>>> amanda.
>>>
>>>The estimate phase on localhost (my backup server) took two hours.

First, don't use 'localhost', even if its the same machine, use the 
FQDN of that machine.  Using localhost will typically come back and 
bite you at recovery time.

>>>  I watched the tar process and it took anywhere from 5-30% of the
>>> cpu
>>
>> Look in /tmp/amanda/sendsize*debug on that system.  It'll have
>> timestamped info on each estimate (probably level 0s and level 1s
>> for each fs). Figure out which one is taking so long.  Any errors
>> in the system logs? What type of hardware RAID?  What OS, and what
>> types of FSs?

DLE's that are all on the same disk should all have the same 'spindle'
number.  Its possible that amanda is running 2 or more sessions 
against the same disk, in which case there will be some lost time due 
to thrashing of the disk with uncoordinated seeks to different 
partitions.

Have you run an hdparm -Tt on the slow disks?  On the raid?

>Can the estimate phase of amanda be bypassed or disabled?

Not that I know of.  amanda needs to figure out how much data there is 
to be stored.  This estimate phase is one of amanda's real strengths 
as it makes for very efficient use of the available resources.

>If so, what are the implications?
>
>Jason Edgecombe

-- 
Cheers, Gene
AMD K6-III@500mhz 320M
Athlon1600XP@1400mhz  512M
99.26% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attornies please note, additions to this message
by Gene Heskett are:
Copyright 2003 by Maurice Eugene Heskett, all rights reserved.