The load is not evenly distributed over ISLs [ Edited ]

[Jan Svoboda]

Hello,

customer has 2 FC switches connected with 2 ISLs.

ISLs have the same speed, are not trunked, exchange-based routing policy used, DLS with lossless enabled,

FOS v7.3.1b, one fabric, HW model is SW7800.

portperfshow on ISL ports 0,4 shows that port 4 is 20x more utilized than 0.
0              4
=============================================
5.0m  101.5m

admin> dlsshow

DLS is set with Lossless enabled

E_Port Balance Priority is not set

islshow (hidden customer specific info)
1: 0-> 0 WWN 220 16-220 sp: 4.000G bw: 4.000G TRUNK CR_RECOV
2: 4-> 4 WWN 220 16-220 sp: 4.000G bw: 4.000G TRUNK CR_RECOV

admin> trunkshow (hidden customer specific info)
1: 0-> 0 WWN 220 deskew 15 MASTER

4: 4-> 4 WWN 220 deskew 15 MASTER

I found command dlsset --rebalance, but not tried, other possible step might be to offline/online one ISL a let rebalance it.

If you have similar experience, let me know, how to rebalance the load, ideally online.

Thanks!

Jan

[ralf.schlosser]

Hi Jan,

can look for the portperfshow for a longer time?

Do you have perhaps MAPS with FPI active?

[Jan Svoboda]

Hello Ralf,

MAPS are enabled with legacy bottleneckmon, FPI will be enabled after update FOS to 7.4.1d.

Dashboard reports port 4 is more utilized than 0.

TX(%) 4(30.70) 4(31.00) 4(30.17) 4(27.93) 4(27.85) 4(27.07) 4(33.33)
      0(5.04)  0(3.69)  0(3.71)  0(2.63)  0(2.84)  0(2.72)  0(4.11)

Bottleneckmon reports port 4 congestion bottleneck at midnight due to backup activity.

(Congestion bottleneck on E-Port 4. 10.00 pct. of 30 secs. affected.)

linkcost

Port Cost
-------------------------
0 (E_PORT) 200 (STATIC)
4 (E_PORT) 200 (STATIC)

Regards,

Jan

[Jan Svoboda]

Hello Ralf,

MAPS are enabled with legacy bottleneckmon, FPI will be enabled after update OFS to 7.4.1d.

Dashboard reports port 4 is more utilized than 0.

TX(%) 4(30.70) 4(31.00) 4(30.17) 4(27.93) 4(27.85) 4(27.07) 4(33.33)
      0(5.04)  0(3.69)  0(3.71)  0(2.63)  0(2.84)  0(2.72)  0(4.11)

Bottleneckmon reports port 4 congestion bottleneck at midnight due to backup activity.

Congestion bottleneck on E-Port 4. 10.00 pct. of 30 secs. affected.

linkcost

Port Cost
-------------------------
0 (E_PORT) 200 (STATIC)
4 (E_PORT) 200 (STATIC)

Regards,

Jan

[Liem]

@Jan Svoboda

You may refer to this: https://community.brocade.com/t5/Fibre-Channel-SAN/load-balancing-over-ifls-with-trunking-not-working-on-brocade/td-p/89563

Regards,

Liem

[Thomas.Letsche]

Hi Jan,

is this imbalance seen all the time so also during the day or just during the backup job at night?

TL

[Thomas.Letsche]

If it's only reporting this imbalance during the backup activity it might be normal due to the Exchange Based Routing you are using. A backup exchange can be very large and would stay on one single path. Seen this several times.