John R. Jackson wrote:
> 
> Each disklist entry has an associated "interface" (field five).  This is
> a "pool" that Amanda takes away from and adds back to as it starts and
> ends backups.  When it picks a candidate backup to start, it computes
> an **estimated** bandwidth that will be used, adds it to the current
> value for that interface, and bypasses the backup (picks the next one)
> if the sum would exceed the "use" value (I think there's an end case
> when it's the last backup that can run).

I think I've gotten myself terribly confused on this.  My tape server is
a Linux box with nice fast SCSI drives, SCSI AIT-2 tape and 100Mb
Ethernet going to a 100Mb switch.  It's backing up Linux boxes with 100Mb,
Solaris boxes with 100Mb and Solaris boxes with 10Mb Ethernet interfaces.
It's also backing up a couple of its local drives.
My disklist looks like:
Chris   /               nocomp-root -1 local
Chris   /var            nocomp-root -1 local
Chris   /home           nocomp-user -1 local
Enceedee        /       nocomp-root -1 enet100
Enceedee        /boot   nocomp-root -1 enet100
Enceedee        /usr/local      nocomp-root -1 enet100
Enceedee        /var    nocomp-root -1 enet100
Noc     /               nocomp-root -1 enet100
Noc     /var            nocomp-root -1 enet100
Noc     /mnt/fast1      nocomp-user -1 enet100
Noc     /usr/local2     nocomp-root -1 enet100
Munin   /               nocomp-root -1 enet10
Munin   /var            nocomp-root -1 enet10
Munin   /usr/local      nocomp-root -1 enet10

Local is for the local disks, enet100 is for clients with 100Mb Ethernet
and enet10 is for clients with 10Mb Ethernet.

I know amanda prefers to do the compressing but half my clients are
antique Suns and don't have enought oomph to do their own compressing.
Even though the tape server's a 500MHz Xeon with 256Mb of RAM the backups
run faster having the tape drive do the compressing.

The pertinent parts of my amanda.conf file looks like:
inparallel 12           # maximum dumpers that will run in parallel (max 63)
netusage  40000 Kbps    # maximum net bandwidth for Amanda, in KB per sec
 
tapebufs 80
# A positive integer telling taper how many 32k buffers to allocate.
define interface local {
    comment "a local disk"
    use 10000 kbps
}
 
define interface enet10 {
    comment "10 Mbps ethernet"
    use 600 kbps
}
 
define interface enet100 {
    comment "100 Mbps ethernet"
    use 6000 kbps
}


Have I got everything messed up in my disklist?  Should I simply remove all
the interface entries since the server box has just the one interface?  I
see messages in the amdump files that I get bandwidth limited at times.  I
don't know if I've upped the netusage variable since that. 
-- 
  [EMAIL PROTECTED] - HMC UNIX Systems Manager

Reply via email to