Dotaz: CISCO ling aggregation nefinguje

[dotaz]

Zdravim,
potreboval bych poradit, snazim se udelat bonding ze 2 stoju XXX, YYY, kazdy pripojeny 2 opt.kabely do CISCO (nexus rada 3000). Ale podle iperfu stale dostavam 10G rychlost, ac bych cekal 20G. Neporadite co je v konfiguraci spatne?
CISCO CONFIG:

interface port-channel1
  description xxx-linkagr
  switchport mode trunk
  switchport trunk allowed vlan 2
  flowcontrol receive on
  flowcontrol send on

interface port-channel2
  description yyy-linkagr
  switchport mode trunk
  switchport trunk allowed vlan 2
  flowcontrol receive on
  flowcontrol send on

------------------------------------

interface Ethernet1/45
  description xxx-bond1
  switchport mode trunk
  switchport trunk allowed vlan 2
  flowcontrol receive on
  flowcontrol send on
  channel-group 1 mode active
  no shutdown

interface Ethernet1/46
  description xxx-bond2
  switchport mode trunk
  switchport trunk allowed vlan 2
  flowcontrol receive on
  flowcontrol send on
  channel-group 1 mode active
  no shutdown

interface Ethernet1/47
  description yyy-bond1
  switchport mode trunk
  switchport trunk allowed vlan 2
  flowcontrol receive on
  flowcontrol send on
  channel-group 2 mode active
  no shutdown

interface Ethernet1/48
  description yyy-bond2
  switchport mode trunk
  switchport trunk allowed vlan 2
  flowcontrol receive on
  flowcontrol send on
  channel-group 2 mode active
  no shutdown

BONDING NA STROJICH:

XXX:
cat /proc/net/bonding/bond0
Ethernet Channel Bonding Driver: v3.6.0 (September 26, 2009)

Bonding Mode: IEEE 802.3ad Dynamic link aggregation
Transmit Hash Policy: layer3+4 (1)
MII Status: up
MII Polling Interval (ms): 200
Up Delay (ms): 0
Down Delay (ms): 0

802.3ad info
LACP rate: slow
Aggregator selection policy (ad_select): stable
Active Aggregator Info:
	Aggregator ID: 7
	Number of ports: 2
	Actor Key: 33
	Partner Key: 32768
	Partner Mac Address: f0:7f:06:84:2b:81

Slave Interface: eth2
MII Status: up
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 1
Permanent HW addr: 0c:c4:7a:1e:3e:02
Aggregator ID: 7
Slave queue ID: 0

Slave Interface: eth3
MII Status: up
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 1
Permanent HW addr: 0c:c4:7a:1e:3e:03
Aggregator ID: 7
Slave queue ID: 0



YYY:
cat /proc/net/bonding/bond0
Ethernet Channel Bonding Driver: v3.6.0 (September 26, 2009)

Bonding Mode: IEEE 802.3ad Dynamic link aggregation
Transmit Hash Policy: layer2 (0)
MII Status: up
MII Polling Interval (ms): 200
Up Delay (ms): 0
Down Delay (ms): 0

802.3ad info
LACP rate: slow
Aggregator selection policy (ad_select): stable
Active Aggregator Info:
	Aggregator ID: 3
	Number of ports: 2
	Actor Key: 33
	Partner Key: 1
	Partner Mac Address: f0:7f:06:84:2b:81

Slave Interface: eth2
MII Status: up
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 1
Permanent HW addr: 0c:c4:7a:1e:0a:2c
Aggregator ID: 3
Slave queue ID: 0

Slave Interface: eth3
MII Status: up
Speed: 10000 Mbps
Duplex: full
Link Failure Count: 1
Permanent HW addr: 0c:c4:7a:1e:0a:2d
Aggregator ID: 3
Slave queue ID: 0

IPERF:

YYY:
iperf -s -p 12345
------------------------------------------------------------
Server listening on TCP port 12345
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[  4] local 147.xxx.yyy.zzz port 12345 connected with 147.xxx.yyy.zzz port 57091
[ ID] Interval       Transfer     Bandwidth
[  4]  0.0-60.0 sec  65.6 GBytes  9.39 Gbits/sec


XXX:
iperf -c yyy -p 12345 -t 60
------------------------------------------------------------
Client connecting to yyy, TCP port 12345
TCP window size: 85.3 KByte (default)
------------------------------------------------------------
[  3] local 147.xxx.yyy.zzz port 57091 connected with 147.xxx.yyy.zzz port 12345
[ ID] Interval       Transfer     Bandwidth
[  3]  0.0-60.0 sec  65.6 GBytes  9.39 Gbits/sec

---

---

Odpovědi

[rpajik]

V jednom TCP spojeni dosahnete rychlosti maximalne jednoho portu, takze ten test iperfem neni relevantni.

[nelson]

Řekl bych, že propustnost tohoto bondu je 2x10Gbps - při dvou různých připojeních je možné dosáhnout v součtu 20Gbps, nikoli 20Gbps na jeden přenos.

[hermelin]

Ale takhle LACP preci nefunguje. Neumi udelat z 2x10GB 20GB port vuci jinemu stroji. Kdyz zkusis iperf ze dvou stroju soucasne na treti stroj tak pak by oba meli jet 10GB protoze LACP rozlozi zatez z ruznych klientu pres ruzna rozhrani a na serveru se vyuziji obe sitovky na max. Ale aby z jednoho PC slo diky LACP 20GB na druhy tak to nejde. Tady by to slo udelat pouze tak ze bys nemel ty sitovky v bondu ale normalne klasicky kazda zvlast a rozlozil by si zatez napr tim ze by na kazde sitovce bezela jina sluzba apod.

[dotaz]

aha, diky

[Ivan]

nejsem si jistej, ale mozna by to slo. implicitne se bere zdrojova a cilova MAC adreasa packetu, z toho se udela hash a z toho se urci kterym dratem to potece. LACP se da pouzit i na vyssi vrstve, pak - myslim - se hashuje i cilo portu. zalazi to jen na volbe hashovaciho algoritmu.

[BigWrigley]

S xmit_hash_policy nastavenou na layer3+4 opravdu půjde rozkládat i provoz mezi jednou dvojicí adres, ale pořád jen v případě, že to bude více než jedno spojení.

[dotaz]

Nastavil jsem to xmit_hash_policy tak jsak jste psal, pustil 2 paralelni iperfy na ruznych portech, kazdy bezi cca 9.90 Gbits/sec. Dekuji