Re: How to determine cause of Amanda slowdown?
2002-08-09 09:51:44
Well, so much for the NIC and switch causing the amanda slowdown. I
changed the settings on my NIC last night before the amanda run to
100baseT-FD. However, the dump is still running now, 15 hours later. I'm
going to examining the switch shortly to see if there were many errors
or collisions.
Here's what amstatus said a few minutes ago:
amanda@admin:~ > amstatus DailySet1
Using /var/log/amanda/DailySet1/amdump from Thu Aug 8 18:00:00 EDT
2002
admin://db/c$ 1 380k finished
(18:14:30)
admin://db/e$ 1 10k finished
(18:13:21)
admin://db/f$ 0 7066846k wait for dumping
admin:sda1 1 10k finished
(18:13:37)
admin:sda3 1 3085k wait for dumping
admin:sdb1 0 3970k finished
(18:13:54)
centernet:sda1 0 2975k finished
(18:16:12)
centernet:sda3 0 1564325k finished (8:22:23)
centernet:sdb1 1 1581k finished
(18:14:42)
centernet:sdb2 0 70452k finished
(18:52:25)
centernet:sdc1 1 1k finished
(18:13:22)
cgi:hda1 0 2294k finished
(18:13:56)
cgi:hda3 0 604189k finished (8:39:55)
kzlaptop:hda5 0 4181k finished
(18:16:44)
kzlaptop:hda7 1 1926k finished
(18:16:16)
mailinglists:hda1 0 943k finished
(18:15:37)
mailinglists:hda2 1 7725k finished
(18:18:42)
mailinglists:hda7 1 41k finished
(18:15:24)
virtual:hda1 0 944k finished
(18:13:35)
virtual:hda3 0 1432404k writing to tape
(8:57:43)
www2:sda10 0 [Request to www2 timed
out.]
www2:sda11 0 [Request to www2 timed
out.]
www2:sda5 0 [Request to www2 timed
out.]
www2:sda7 0 [Request to www2 timed
out.]
www2:sda8 0 [Request to www2 timed
out.]
www2:sda9 0 [Request to www2 timed
out.]
SUMMARY part real estimated
size size
partition : 26
estimated : 0 0k
failed : 6 0k ( 0.00%)
wait for dumping: 2 7069931k ( 0.00%)
dumping to tape : 0 0k ( 0.00%)
dumping : 0 0k 0k ( 0.00%) ( 0.00%)
dumped : 18 3698351k 4633924k ( 79.81%) ( 0.00%)
wait for writing: 0 0k 0k ( 0.00%) ( 0.00%)
writing to tape : 1 1432404k 1342192k (106.72%) ( 0.00%)
failed to tape : 0 0k 0k ( 0.00%) ( 0.00%)
taped : 17 2265947k 3291732k ( 68.84%) ( 0.00%)
8 dumpers idle : no-diskspace
taper writing, tapeq: 0
network free kps: 2600
holding space : 2062091k ( 59.01%)
dumper0 busy : 14:14:54 ( 96.67%)
dumper1 busy : 13:38:37 ( 92.57%)
dumper2 busy : 0:05:06 ( 0.58%)
dumper3 busy : 14:44:22 (100.00%)
dumper4 busy : 0:03:14 ( 0.37%)
dumper5 busy : 0:00:59 ( 0.11%)
taper busy : 0:42:07 ( 4.76%)
0 dumpers busy : 0:00:00 ( 0.00%)
1 dumper busy : 0:28:59 ( 3.28%) no-bandwidth: 0:28:59
(100.00%)
2 dumpers busy : 0:36:32 ( 4.13%) no-bandwidth: 0:36:32
(100.00%)
3 dumpers busy : 13:33:43 ( 92.01%) no-bandwidth: 13:33:42
(100.00%)
4 dumpers busy : 0:02:06 ( 0.24%) no-bandwidth: 0:01:54 (
90.50%)
start-wait: 0:00:12 (
9.50%)
5 dumpers busy : 0:02:30 ( 0.28%) no-bandwidth: 0:02:03 (
82.11%)
client-constrained: 0:00:25 (
17.09%)
start-wait: 0:00:01 (
0.80%)
6 dumpers busy : 0:00:30 ( 0.06%) no-diskspace: 0:00:15 (
49.47%)
no-bandwidth: 0:00:12 (
40.64%)
client-constrained: 0:00:02 (
6.82%)
amanda@admin:~ >
In the first section, the admin://db/f$ is a samba connection to a
Windows NT host through the tapebackup host. The F: drive is primarily
one huge database file. The admin:sda3 is the root directory on the
tapebackup host itself. I should have commented out www2, since I know
it's a dead host.
What's the meaning of this line in the second section: "8 dumpers idle
: no-diskspace" Is this an error message? Should I try to allocate more
disk space to the dump disk?
Do the last lines about 1-6 dumpers busy, due to no-bandwidth indicate
that I need to increase the netusage? I currently have "netusage 1200
Kbps # maximum net bandwidth for Amanda, in KB per sec" Since I'm on
a 100baseT-FD line to the switch, can I increase this? Is my thinking
correct that I'm on a 100Mbps line, and have set the maximum bandwidth
usage to 1.2Mbps? If this is true, I would think I can surely increase
it. What's a prudent value, 25% of the line speed, 50%, 100%? The
backups run at night, where there's very little activity on our LAN.
Does the peak at 3 dumpers busy have any significance?
Thank you all again for your help and suggestions.
-Kevin
|
|
|