[LARTC] FW: QoS extension to Net-SNMP

[LARTC] FW: QoS extension to Net-SNMP

[Heywood, Simon]

Hi.

I did send this to `jaazz@post.cz', but I suspect the list is a more
appropriate/useful place for it. It's a question about Michal Charvat's QoS
extension to Net-SNMP.

When I look at the MIB entries for the QoS handles, I get something like
this -

enterprises.qos.qosObjectTable.qosObject.qosHandle.0.0 = Gauge32: 0
enterprises.qos.qosObjectTable.qosObject.qosHandle.5.65536 = Gauge32: 65536
enterprises.qos.qosObjectTable.qosObject.qosHandle.5.131072 = Gauge32:
131072
enterprises.qos.qosObjectTable.qosObject.qosHandle.5.131073 = Gauge32:
131073
enterprises.qos.qosObjectTable.qosObject.qosHandle.5.131088 = Gauge32:
131088
enterprises.qos.qosObjectTable.qosObject.qosHandle.5.131089 = Gauge32:
131089
enterprises.qos.qosObjectTable.qosObject.qosHandle.5.131090 = Gauge32:
131090
enterprises.qos.qosObjectTable.qosObject.qosHandle.5.5308416 = Gauge32:
5308416

Now, mapping these handles to major:minor numbers I get -

0:0
1:0
2:0
2:1
2:16
2:17
2:18
81:0

I did this using the following calculation (in Java, FWIW) -

major = handle / 0x00010000L
minor = handle % 0x00010000L


However, tc gives a different story -

$ tc -s -d qdisc show dev sync1
qdisc pfifo 51: limit 5p
 Sent 0 bytes 0 pkts (dropped 0, overlimits 0) 
qdisc htb 2: r2q 10 default 12 direct_packets_stat 0 ver 3.10
 Sent 7279652 bytes 215480 pkts (dropped 0, overlimits 0) 
qdisc dsmark 1: indices 0x0008 default_index 0x0003 set_tc_index 
 Sent 7279652 bytes 215480 pkts (dropped 0, overlimits 0) 


$ tc -s -d class show dev sync1
class htb 2:11 parent 2:1 leaf 51: prio 2 quantum 12500 rate 125000bps ceil
125000bps burst 15Kb/8 mpu 0b cburst 2849b/8 mpu 0b level 0 
 Sent 0 bytes 0 pkts (dropped 0, overlimits 0) 
 lended: 0 borrowed: 0 giants: 0
 tokens: 100663 ctokens: 18677

class htb 2:1 root rate 250000bps ceil 250000bps burst 15Kb/8 mpu 0b cburst
4Kb/8 mpu 0b level 7 
 Sent 7279664 bytes 215481 pkts (dropped 0, overlimits 0) 
 rate 15bps 
 lended: 0 borrowed: 0 giants: 0
 tokens: 50305 ctokens: 13408

class htb 2:10 parent 2:1 prio 1 quantum 25000 rate 250000bps ceil 250000bps
burst 15Kb/8 mpu 0b cburst 4Kb/8 mpu 0b level 0 
 Sent 0 bytes 0 pkts (dropped 0, overlimits 0) 
 lended: 0 borrowed: 0 giants: 0
 tokens: 50331 ctokens: 13434

class htb 2:12 parent 2:1 prio 3 quantum 6250 rate 62500bps ceil 250000bps
burst 15Kb/8 mpu 0b cburst 4Kb/8 mpu 0b level 0 
 Sent 7279664 bytes 215481 pkts (dropped 0, overlimits 0) 
 rate 12bps 
 lended: 215481 borrowed: 0 giants: 0
 tokens: 201222 ctokens: 13408


$ tc -s -d filter show dev sync1
filter parent 1: protocol ip pref 1 tcindex hash 0 mask 0x00fc shift 2
pass_on 


In other words, the handles don't seem to match those given by tc (which are
the same handles specified in the config. script).

Have I done something wrong...?

Thanks for your time.

S.
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

Re: [LARTC] FW: QoS extension to Net-SNMP

[Stef Coene]

On Friday 30 January 2004 10:59, Heywood, Simon wrote:
> Hi.
>
> I did send this to `jaazz@post.cz', but I suspect the list is a more
> appropriate/useful place for it. It's a question about Michal Charvat's QoS
> extension to Net-SNMP.
<snip>

> In other words, the handles don't seem to match those given by tc (which
> are the same handles specified in the config. script).
>
> Have I done something wrong...?
I don't know.  But I did the same in perl.  And my results are ok.  You can 
find the scripts on docum.org.
http://docum.org/stef.coene/qos/tc-snmp/index.html

Stef

-- 
stef.coene@docum.org
 "Using Linux as bandwidth manager"
     http://www.docum.org/
     #lartc @ irc.openprojects.net

_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/

RE: [LARTC] FW: QoS extension to Net-SNMP

[Heywood, Simon]

Michal Charvat wrote:
> But as I see yours output.... I have one question. Do you have x86
> platform? I didn't try that on other than x86 and there can be
> problem with __u32 interpretation. 

No, it's all x86.

Anyway, I think I've solved the problem - the numbers I've got out of SNMP
and converted to major:minor values are in base 10, whereas tc apparently
uses hexadecimal for class IDs. Of course, no-one bothered to tell me this
when I was introduced to tc. :-)

(Thanks to Stef Coene for unwittingly pointing this out to me via his
scripts.)

S.
_______________________________________________
LARTC mailing list / LARTC@mailman.ds9a.nl
http://mailman.ds9a.nl/mailman/listinfo/lartc HOWTO: http://lartc.org/