RE: Average Dump Rate Question
Check your hme interface settings. I found that on the systems that
appeared to be slow, the NIC was configured for something other than 100
full, even though the physical network was 100 full.
If you have access to SunSolve ( http://sunsolve.sun.com/ ), you can check
out FAQ ID #2605 and Infodoc ID #18262 for some very useful information, as
well as some good instructions on changing your settings. You'll probably
have to have a SunSolve account to get the info. If you've read and
understood the docs, then the following might be helpful (at least it can
show you if the settings are conflicting):
To see the list of variables for the /dev/hme interface, use:
# ndd /dev/hme \?
I use this little script to show me the current settings:
#
VARLIST=`ndd /dev/hme \? | sed '1d' | cut -d" " -f1`
for VAR in $VARLIST
do
echo "${VAR} = `ndd /dev/hme ${VAR}`"
done
#
The output looks like this: (some comments are added in with #)
# ./nddcheck.sh
transceiver_inuse = 0
link_status = 1 # 1=up 0=down (Read-Only)
link_speed = 1 # 1=100mps 0=10mps (R-O)
link_mode = 0 # 1=full duplex 0=half duplex (R-O)
ipg1 = 8
ipg2 = 4
use_int_xcvr = 0
pace_size = 0
adv_autoneg_cap = 1 # Advertise whether you want to
auto-negotiate
# If adv_autoneg_cap = 0,
set one of the following
# adv_values to 1 and all
the others to 0.
# If adv_autoneg_cap = 1,
then
adv_100T4_cap = 0 # Advertise available
speeds/modes
adv_100fdx_cap = 1 # 100 full
adv_100hdx_cap = 1 # 100 half
adv_10fdx_cap = 1 # 10 full
adv_10hdx_cap = 1 # 10 half
autoneg_cap = 1 # Current settings (read-only
values)
100T4_cap = 0
100fdx_cap = 1
100hdx_cap = 1
10fdx_cap = 1
10hdx_cap = 1
lp_autoneg_cap = 0 # Link partner settings (read-only
values)
lp_100T4_cap = 0
lp_100fdx_cap = 0
lp_100hdx_cap = 0
lp_10fdx_cap = 0
lp_10hdx_cap = 0
instance = 0# /dev/hme0
lance_mode = 1
ipg0 = 16
So according to the settings, this system is configured to auto-negotiate
and has settled on 100 Mbps, half duplex. Auto-negotiate slows things down,
and so does the half duplex...
Check with the people who configure your network/routers/switches to be sure
of the settings you need to use. Also, if the machine is on a hub, that too
should be taken into consideration.
Hope this helps...
David Carter
McLeodUSA Information Systems
[EMAIL PROTECTED]
281-465-1835
> -Original Message-
> From: Tanniel Simonian [mailto:[EMAIL PROTECTED]]
> Sent: Saturday, March 03, 2001 4:11 PM
> To: [EMAIL PROTECTED]
> Subject: Average Dump Rate Question
>
>
>
> Currently my average dump rate on a 100mbit full switched network
> consisting of Cisco 2948G's and Dell Pentium III 800 mhz with
> 512 megs of
> ram server is at:
>
> Avg Dump Rate (k/s) 364.5 364.5--
>
> How can I speed this up?
> In amanda.conf I have set netusage to 5 kbits.
> should I increase that number, decrease the number of dumps running
> concurrently, shouldn't my network card be able to handle at least 4
> megabytes per second, (3com 35905B 100mbit)? Does high
> compression play a
> factor in the file transfer, or does the compression be done
> on the client
> machine then sent over the network to the dump disk?
>
> My hard disk also is set up as this:
> define interface local {
> comment "a local disk"
> use 5 kbps
> }
>
> My local disk is an IBM Ultra 2 9 gig 1 RPM drive. I believe 5
> kbps is sort of slow for a drive like this, but I don't get any
> improvements by bumping this number up?
>
> What can I do to increase my dump rate? Also does amanda ignore, when
> defining a tape type the speed of what the tape drive is
> capable? Does it
> leave determining the speed to the actual drive and interface?
>
> define tapetype M2-AME225 {
> comment "Exabyte EZ17 Mammoth2"
> length 57487 mbytes
> filemark 0 kbytes
> speed 11765 kbytes
> }
>
> Avg Tp Write Rate (k/s) 6689.9 6689.9--
>
> Im assuming that compression may play a role in the speed of the tape
> drive copying content?
>
> Any suggestions are greatly appreciated,
>
> Thank you,
>
> Tanniel Simonian
> University of California Riverside Science Library
> Linux Admin/Networking Admin
>
>
>
Re: A question about chg-zd-mtx
Rock on! I complained about that a while ago. Since 1.2.10 uses CHANGER, I would set chg-zd-mtx to use CHANGER and say that it requires 1.2.10 or above. --Joe, running to update from 1.2.9 to 1.2.10 ;-) On Sat, Mar 03, 2001 at 12:43:15AM -0500, Jason Hollinden wrote: > Now I'm more confused. :) > > The docs for mtx-1.2.10 say to use CHANGER, while the older versions > seemed to use TAPE. I assume that's why the chg-zd-mtx was written > with the $TAPE variable. 1.2.10 also will honor TAPE or CHANGER, which was > why I had never changed it before. > > If if it doesn't matter if it's called TAPE or CHANGER, I'd like to have > it as CHANGER, since that what their docs now suggest. > > Thoughts? > > On Fri, 02 Mar 2001, Joe Rhett wrote: > > > Sorry - wrong maintainer. The problem isn't chg-zd-mtx, it's the 'mtx' > > program itself which uses the $TAPE environment variable. Sorry for the > > confusion. > > > > On Fri, Mar 02, 2001 at 09:22:44AM -0500, Jason Hollinden wrote: > > > Here is a patch for the current CVS version of chg-zd-mtx.sh.in that > > > changes the $TAPE variable to $CHANGER. It was only used 3 times, so it > > > was pretty painless. > > > > > > On Thu, 01 Mar 2001, Joe Rhett wrote: > > > > > > > As others have said, it uses the TAPE environment variable. I wish we could > > > > convince the maintainer to use CHANGER instead, so it doesn't conflict with > > > > mt's TAPE environment variable > > > > > > > > Anyway, make sure you are using either the version we made > > > > http://www.noc.isite.net/?Projects > > > > > > > > or the version someone else did which supports barcodes. The version > > > > distributed with Amanda doesn't work with modern versions of mtx and has > > > > numerous bugs. > > > > > > > > On Wed, Feb 28, 2001 at 01:24:12PM -0500, Stan Brown wrote: > > > > > I'm trying to get my first tape changer working, and I'm a little confused. > > > > > > > > > > It appears that I need to use the chg-zd-mtx scriptm but the version of mtx > > > > > that I have is used like this: > > > > > > > > > > mtx -f /dev/sg0 next > > > > > > > > > > for instance. Now it appears to me that chg-zd-mtx just runs mtx like this: > > > > > > > > > > mtx next > > > > > > > > > > That is wihtout the device specifier. Now it is possible for my version of > > > > > mtx to use the CHANGER environment variable, instead of the -f argument, > > > > > but I can't see anywhere that gets set in chg-zd-mtx either, and since this > > > > > will ultimatley be run fron cron, and thus potentialy be somewhat > > > > > "environmently chalenged", I am wondering how this is usually handled? > > > > > > > > > > > > > > > -- > > > > > Stan Brown [EMAIL PROTECTED]843-745-3154 > > > > > Charleston SC. > > > > > -- > > > > > Windows 98: n. > > > > > useless extension to a minor patch release for 32-bit extensions and > > > > > a graphical shell for a 16-bit patch to an 8-bit operating system > > > > > originally coded for a 4-bit microprocessor, written by a 2-bit > > > > > company that can't stand for 1 bit of competition. > > > > > - > > > > > (c) 2000 Stan Brown. Redistribution via the Microsoft Network is prohibited. > > > > > > > > -- > > > > Joe Rhett Chief Technology Officer > > > > [EMAIL PROTECTED] ISite Services, Inc. > > > > > > > > PGP keys and contact information: http://www.noc.isite.net/Staff/ > > > > > > > > > -- > > >Jason Hollinden > > > > > >SMG Systems Admin > > > > > --- chg-zd-mtx.sh.in.20010302 Fri Mar 2 08:57:17 2001 > > > +++ chg-zd-mtx.sh.in Fri Mar 2 09:00:39 2001 > > > @@ -88,8 +88,8 @@ > > > > > > myname=$0 > > > tape=`amgetconf$SUF tapedev` > > > -TAPE=`amgetconf$SUF changerdev`; export TAPE # for mtx command > > > -if [ "$tape" = "/dev/null" -o "$TAPE" = "/dev/null" ]; then > > > +CHANGER=`amgetconf$SUF changerdev`; export CHANGER # for mtx command > > > +if [ "$tape" = "/dev/null" -o "$CHANGER" = "/dev/null" ]; then > > > echo "Both tapedev and changerdev must be specified in config file"; > > > exit 2; > > > fi > > > > > > -- > > Joe Rhett Chief Technology Officer > > [EMAIL PROTECTED] ISite Services, Inc. > > > > PGP keys and contact information: http://www.noc.isite.net/Staff/ > > > -- >Jason Hollinden > >SMG Systems Admin -- Joe Rhett Chief Technology Officer [EMAIL PROTECTED] ISite Services, Inc. PGP keys and contact information: http://www.noc.isite.net/Staff/
Re: Do I get this right ?
>I'm in new in this amanda world. Welcome! >First the cron job on the server launches the amdump >The amanda checks the disklist and connect to the client via UDP >The Client answer back via UDP and launches amandad which takes care of >the transit. >Is it right ? Pretty much, although the actual data transfer of the backup image, error messages and catalogue (if enabled) is done with TCP. It goes like this: Amanda Server Amanda Client = = send UDP request to client:10082 inetd sees UDP request and starts amandad amandad reads and decodes packet, performs security checks and starts the requested "service" (selfcheck, sendsize, sendbackup) sendbackup creates two or three TCP sockets (data, messages, index/catalogue) and sends those port numbers back to the server in a UDP packet client waits for incoming connections on the new ports get port numbers from client and connect via TCP accepts connections and transfers data >One other question, can soemone tells me which port I have to open on the >client side ? UDP 10080 is the only thing a client needs. The server needs TCP 10082 and 10083 if you plan on using amrecover. When the sendbackup TCP sockets are created on the client and when the corresponding sockets are created on the server, ports are chosen first from the range you gave ./configure with --with-portrange (if you set that at all), then privileged ports (512 .. 1023) are tried, then any available port. When programs are not running as root (such as sendbackup), they cannot get a privileged port, so that part does not apply. UDP ports are bound to specific ranges just like TCP ports. Amcheck, planner and dumper use that so you can limit the values selected to get through a firewall. >I have removed the UDP 10080, TCP 10082 - 10083 from the >"services" file on the client machine, restarted inetd and amcheck still >works. Is it normal ? I don't know how that could work. After removing those entries from your services file, inetd should have been unhappy with your configuragion file and not accepted connections. There must be something else going on here. >Olivier Collet John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Re: Help with Disk Drive Backup
> I am new at using amanda. Where could I find help on using a spare disk >drive on another machine for backup. ... Do you mean use a disk instead of a tape? You can do this a couple of ways. First, you can set up the disk as your holding disk and set tapedev to "/no-such-device" (or anything else that does not exist). Amanda will do all the dumps into the holding disk and just never write them to tape. You also need to change the "reserve" holding disk parameter (see amanda(8)) so it will do full dumps. And you'll need to clean out old dumps by hand. The other way is to get the amanda-242-tapeio CVS branch and use the new "file:" tape driver. That lets Amanda work normally without all the problems of only using a holding disk and treats the extra disk like a tape. The "file:" driver is described in the version of amanda(8) you'll get with the CVS branch. Note that this code is somewhat experimental. >Luis. John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED]
Help with Disk Drive Backup
Hello, I am new at using amanda. Where could I find help on using a spare disk drive on another machine for backup. I.E. using NFS and/or using SAMBA. Is there an easy way to do this? Thanks for any help ahead of time, Luis.
Do I get this right ?
Hello, I'm in new in this amanda world. I just wanted to know If I figured out correctly what happend exactly when amanda does a backup. First the cron job on the server launches the amdump The amanda checks the disklist and connect to the client via UDP The Client answer back via UDP and launches amandad which takes care of the transit. Is it right ? One other question, can soemone tells me which port I have to open on the client side ? I have removed the UDP 10080, TCP 10082 - 10083 from the "services" file on the client machine, restarted inetd and amcheck still works. Is it normal ? Thanks in advance Olivier Collet
Re: Backup tape issue
* John R. Jackson <[EMAIL PROTECTED]> (Sat, Mar 03, 2001 at 10:31:14AM -0500) > >With that all said, he would like to be able to put ANY tape in the drive > >and have it succeed, basicly ignoring the tape lables. ... > > Then I assume you're not going to be using incrementals but doing a full > dump of everything every day, right? Because this won't work if there are > incrementals. You'll end up clobbering something you need for a restore. > > In the extreme case, your boss wants to be able to use the same tape > every day. To do that with Amanda, set dumpcycle to 0 and tapecycle > to 1. You may have more tapes listed in tapelist (via amlabel) and, > as I understand it, Amanda will accept any of them. But you are seriously fucked if you have only one tape, and the disk you're backing up dies while writing the current backup over the previous one. I suggest a (smart-)moneky proof system, simply labelling the tapes *MONDAY* *TUESDAY* .. *FRIDAY* and go with that. Setting dumpcycle to 0 and tapecycle to 1 will ensure that it will write on whatever tape is in the drive. But clearly labelling them, and training people to put thet tape with the current day in the tapedrive should minimize the risk of the above scenario. > > >Richard G. Duvall > > John R. Jackson, Technical Software Specialist, [EMAIL PROTECTED] > Currently listening to: Godspeed you black emperor - Levez vos skinny fists.. CD1-2 Gerhard, <@jasongeo.com> == The Acoustic Motorbiker == -- __O Get one man to ride his bycicle =`\<, and there is hope for the human race (=)/(=) A.A.M. van Gulik
Re: amrestore problem
vi /etc/services /amandaidx if you know how to use vi you can figure it out from there. otherwise: man vi w >> >amandaidx stream tcp nowait amanda /usr/sbin/tcpd/ usr/local/libexec >> >amidxtape stream tcp nowait amanda /usr/sbin/tcpd /usr/local/libexe >> >amanda dgram udp wait amanda /usr/sbin/tcpd /usr/local/libexec >> Those cannot be right. First, there is a trailing '/' on "tcpd" for >> amandaidx. Next, all the lines end with "libexec", which is a directory, >> not a program. > >Hm...and what is the right way to change this parameters? >Can you give me a tip? > >Bye Jeurgen > > >-- >Dies ist eine Microsoft freie Mail!
