I've noticed that the "inparallel" setting appears to not be honored. The
server is running 2.6.1p2, redhat enterprise 5.4, it has a 300GB holding
disk and writes to disk (vtapes). Here are some relevant portions of
amanda.conf:
org "daily" # your organization name for
reports
dumpuser "amanda" # the user to run dumps
under
inparallel 15 # maximum dumpers that will run in
parallel (max 63)
maxdumps 2
dumporder "SSSSSSSSSSSSSSS" # specify the priority order of each
dumper
taperalgo largest # The algorithm used to choose which dump
image to send
displayunit "m" # Possible values: "k|m|g|t"
netusage 750000 Kbps # maximum net bandwidth for Amanda, in KB
per sec
dumpcycle 7 days # the number of days in the normal dump
cycle
runspercycle 0 # the number of amdump runs in dumpcycle
days
runtapes 10 # number of tapes to be used in a single run
of amdump
holdingdisk hd1 {
directory "/hold/amandaholdingdisk" # where the holding disk is
use -35Gb # how much space can we use on it
chunksize 10Gb # size of chunk if you want
big dump to be
}
define tapetype vtape {
comment "Amanda Virtual Tape, capacity 150G"
length 155000 mbytes
}
define dumptype global {
program "GNUTAR"
index yes
record yes
maxdumps 2
holdingdisk yes
tape_splitsize 10 Gb
fallback_splitsize 2 Gb
compress none
auth "ssh"
ssh_keys "/home/amanda/.ssh/id_rsa_amdump"
}
define interface eth0 {
comment "1000 Mbps ethernet"
use 700000 kbps
}
When I look at the DLEs being dumped with amstatus I see:
Using /usr/local/amanda/logfiles/amdump
>From Tue Dec 8 20:05:01 CST 2009
server1.utexas.edu:/home 0 66880m dumping 66880m
(100.00%) (20:44:22) (waiting for holding disk space)
server2.utexas.edu:/home 1 57426m dumping 51399m (
89.50%) (20:44:22)
server3.utexas.edu:/ 1 43605m dumping 43605m
(100.00%) (20:44:22) (waiting for holding disk space)
At most I've seen 6 items being dumped at any one time. The load average on
the server is less than 1 and five minute avg on the NIC is 70Mbps. There's
over 80-80GB left on the holding disk and all dumps are set to use the
holding disk. I found a previous message indicating issues with inparallel
but that was fixed by upping the netusage option. I increased this setting
but still only see a few DLEs being dumped.
Any idea what's going on?
By the way, the "waiting for holding disk space" is another question ...
especially since there is more than sufficient holding space.
