Amanda-Users

Re: concurrancy

2004-01-22 13:54:12
Subject: Re: concurrancy
From: Gene Heskett <gene.heskett AT verizon DOT net>
To: amanda-users AT amanda DOT org
Date: Thu, 22 Jan 2004 13:49:32 -0500
On Thursday 22 January 2004 10:11, Brian Cuttler wrote:
>Stefan,
>Gene,
>Paul,
>
>Unfortunately my amanda run didn't run any more quickly.
>
>>From amanda.conf - I can send more if you want to see more.
>
>org "NOTES_DLT"         # your organization name for reports
>mailto "amanda-adm AT newton.wadsworth DOT org"        # the mailing list
> for operators at your site
>dumpuser "bin"          # the user to run dumps under

I would hope that the backups are being done as an unprivileged user, 
bin has quite high rights on most systems.  OTOH, the mailto: here is 
root.

Thats why most of us have added a user 'amanda' and made amanda a 
member of the group disk or bin, whichever works for your distro.

># changed from 8 to 4... can't explain excessive time to backup 4
> partitions -ck inparallel 4            # maximum dumpers that will
> run in parallel # maxdumps, added 21-jan-2004 BRC
>maxdumps   2            # max (concurrent) to be run on any
> individual "client"
>
>disklist
>--------
># Notes server backup
>wcnotes /             comp-root
>wcnotes /maildb       comp-user
>wcnotes /maildb2      comp-user
>wcnotes /export/home  comp-user

Unless you have an exclude list file in comp-root that 
specifies ./maildb, ./maildb2 and ./home (on separete lines) you are 
doing the whole of the last 3 filesystems twice each, once for / and 
once per subdir listing.  The last entry will also make you require a 
seperate disklist entry for /home that doesn't specify the exclude 
file.

Thats of course unless the last 3 are in fact different disks 
entirely, in which case you should append a 1, a 2, a 3, or a 4 in 
order to define to amanda that they are separate physical disks and 
can therefore be done concurrently.  One should not attempt to backup 
2 different partitions on the same disk concurrently as that will 
thrash the seeks, slowing things down a bit, or potentially 
overheating the seek drivers and damaging the drive, but that would 
be a very extreme result and something I've not seen in at least 5 
years.

I have the situation in my /usr dir that there are a couple of the 
subdirs that are more than a tapefull, so my /usr is broken down into 
a /usr/subdir per disklist entry.  The is no entry for / in my 
disklist.

There are also good reasons to use a non-compressing dumptype for some 
dirs as they will grow rather than compress if they contain mostly 
executable binaries.  Likewise for dirs that contain already 
compressed archives & rpms.  Its a waste of gzips time to try and 
compress them further.  A good argument for a bit of reason in the 
directory layouts :)

You can see which ones you are wasting compression efforts (and cpu 
horsepower) on by looking at the reported compression achieved, in 
the email from amanda, and switching those entrys that only get 5 or 
10% to a non-compressing dumptype, there by saveing a lot of useless 
wheel-spinning by gzip.

>that is it, just the one client. Suppose I can set maxdumps to 4,
> anyway no sense in having inparallel != maxdumps with only a single
> client.
>
>How sensitive is amanda.conf (disklist or others) to white space ?
>I've notices that its often

It doesn't seem to be at all sensitive to whitespace here.  I use tabs 
liberally for formatting in mine, but an equal amount of spaces is 
also just fine.

>keyword   value <tab># comment
>
>Just to be paraniod I changed the space to a tab.
>
>Am I missing some other parameter that controls parallelism ?

The use of spindle numbers in the dle perhaps.

Here are a few entrys from mine to illustrate:
---
coyote  /var                    comp-root-tar   1       local
coyote  /usr/bin                root-tar        1       local
gene    /boot                   root-tar        2       le0
gene    /home                   comp-root-tar   2       le0
---
where the final local or le0 tells anamda that the disk is local or 
out on the network.  Also, the /etc/hosts files on both machine 
contain the alias<->FQDN of the machines, hence the shortand here is 
usable.

>I was thinking of running amstatus from cron very 15 minutes
>to watch it but that still isn't much of a diagnostic, more to
>confirm lack of parallelism.
>
>                                               thanks,
>
>                                               Brian
>
>---
>   Brian R Cuttler                 brian.cuttler AT wadsworth DOT org
>   Computer Systems Support        (v) 518 486-1697
>   Wadsworth Center                (f) 518 473-6384
>   NYS Department of Health        Help Desk 518 473-0773

-- 
Cheers, Gene
"There are four boxes to be used in defense of liberty: soap,
ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
99.22% setiathome rank, not too shabby for a WV hillbilly
Yahoo.com attornies please note, additions to this message
by Gene Heskett are:
Copyright 2004 by Maurice Eugene Heskett, all rights reserved.

<Prev in Thread] Current Thread [Next in Thread>