Fibre Channel (SAN)

Reply
Contributor
Posts: 21
Registered: ‎04-08-2016

Congestion Bottleneck -- No frame disc

[ Edited ]

Hi

 

Am looking for a solution, to find why my port is loging congestion bottleneck when there is no frames are getting disc in this port and no other soft errors like CRC , link fail and nothing.

 

Meanwhile i checked portperfshow i could see the below output:

>portperfshow 11/3
          3       Total
========================
slot 11:   1.2g   1.2g
          3       Total
========================
slot 11:   1.2g   1.2g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.2g   1.2g
          3       Total
========================
slot 11:   1.2g   1.2g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.3g   1.3g
          3       Total
========================
slot 11:   1.2g   1.2g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.2g   1.2g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.1g   1.1g
          3       Total
========================
slot 11:   1.2g   1.2g

 

 

Can someone tell me whether this is acceptable for 8gbs port?

 

Thanks in advance!

Valued Contributor
Posts: 555
Registered: ‎03-20-2011

Re: Congestion Bottleneck -- No frame disc

You'd better do portperfshow with tx rx flags, because this way we can only see the sum of both. But it's more or less obvious that one of the directions is overloaded and you probably need to add one more link here.
Contributor
Posts: 21
Registered: ‎05-26-2010

Re: Congestion Bottleneck -- No frame disc

Brocade bottleneck "congestion" alerts does not mean there will be dropped frames.  Congestion is an indicator the link has approach limit b/w capacity.  Latency alerts are more likely to be associated with dropped frames.  Compare your link speed vs. link utilization levels. 

Brocadian
Posts: 1
Registered: ‎08-09-2010

Re: Congestion Bottleneck -- No frame disc

HI,

Do a portstatsshow on 11/3. Are the bb credit zero incrementing? You can get a bottleneck hit just from bb credit zero incrementing and no class 3 discards. See hat ports are talking to this port and any associated ISLs and do portstatsshow on those ports too.

Contributor
Posts: 21
Registered: ‎04-08-2016

Re: Congestion Bottleneck -- No frame disc

Alexey: thanks for the response.

 

I tried with tx rx parameter where i could see the tx is more than rx please find the output below.

 

portperfshow 11/3 -tx  -rx
          3
=======================
slot 11: 423.1m  97.5m
          3
=======================
slot 11: 541.2m  74.1m
          3
=======================
slot 11: 504.9m 101.8m
          3
=======================
slot 11: 528.5m  54.1m
          3
=======================
slot 11: 564.6m  90.7m
          3
=======================
slot 11: 582.5m  66.5m
          3
=======================
slot 11: 758.6m 392.1m
          3
=======================
slot 11: 715.9m 470.4m
          3
=======================
slot 11: 418.5m 180.9m
          3
=======================
slot 11: 719.2m 271.9m
          3
=======================
slot 11: 625.9m 205.6m
          3
=======================
slot 11: 548.0m 117.7m
          3
=======================
slot 11: 548.0m 117.7m
          3
=======================
slot 11: 546.4m 123.0m
          3
=======================
slot 11: 580.0m 108.5m
          3
=======================
slot 11: 479.8m 190.0m
          3
=======================
slot 11: 514.5m 156.0m
          3
=======================
slot 11: 461.9m 181.0m
          3
=======================
slot 11: 324.3m 162.7m
          3
=======================
slot 11: 554.3m 134.2m
          3
=======================
slot 11: 601.7m 165.6m
          3
=======================
slot 11: 691.9m 194.0m
          3
=======================
slot 11: 732.2m 218.0m
          3
=======================
slot 11: 684.6m 176.8m
          3
=======================
slot 11: 709.3m 191.8m
          3
=======================
slot 11: 749.4m 191.1m
          3
=======================
slot 11: 759.5m 179.3m
          3
=======================
slot 11: 682.3m 186.9m
          3
=======================
slot 11: 781.0m 206.2m

 

 

The 11/3 is 8gbps port is connected to IBM SVC DH8 ( 4 node cluster ) Node4's P1 . I do tried to see the  SVC Node4's p3 perfshow just to check how its throughput looks. Please have a look at it below.

 

DCR_A:FID128:kalaadmin> portperfshow 12/3 -tx -rx
=======================
slot 12:  25.5m 328.0m
          3
=======================
slot 12:  34.3m 295.2m
          3
=======================
slot 12:  19.1m 342.5m
          3
=======================
slot 12:  12.6m 339.5m
          3
=======================
slot 12:  14.4m 235.9m
          3
=======================
slot 12:  20.2m 242.1m
          3
=======================
slot 12:  15.3m 264.2m
          3
=======================
slot 12:  11.5m 296.8m
          3
=======================
slot 12:  12.9m 286.0m
          3
=======================
slot 12:  14.9m 286.4m
          3
=======================
slot 12:  13.5m 301.2m
          3
=======================
slot 12:  14.2m 304.2m
          3
=======================
slot 12:  26.2m 241.0m
          3
=======================
slot 12:  22.9m 274.7m
          3
=======================
slot 12:  31.5m 324.8m
          3
=======================
slot 12:  27.1m 320.6m
          3
=======================
slot 12:  10.1m 278.7m
          3
=======================
slot 12:  15.9m 306.1m
          3
=======================
slot 12:  11.8m 250.9m
          3
=======================
slot 12:  19.3m 364.9m
          3
=======================
slot 12:  16.4m 271.2m
          3
=======================
slot 12:  19.8m 276.8m
          3
=======================
slot 12:  14.4m 254.2m
          3
=======================
slot 12:  19.1m 256.2m
          3
=======================
slot 12:  20.2m 269.5m
          3
=======================
slot 12:  15.7m 255.9m

 

Not sure why 11/3 tx is high. Is there a way to drill down why this is happening for this port all of a sudden from past few weeks?

 

As you mentioned " need to add more link here " could you please little elaborate it?

Contributor
Posts: 21
Registered: ‎04-08-2016

Re: Congestion Bottleneck -- No frame disc

nguyenb: Thanks for your response.

 

I did portstatsshow, where i dont see any type of soft errors all parameters are clean and clear. I too expected there would be bb credit zero increments, but it wasn't there.

Contributor
Posts: 21
Registered: ‎04-08-2016

Re: Congestion Bottleneck -- No frame disc

[ Edited ]

dtocci: Thanks for you response.

 

I go with your comments.

 

While i notice my output porperfshow 11/3 -tx -rx

 

portperfshow 11/3 -tx -rx
          3
=======================
slot 11: 700.8m  96.2m
          3
=======================
slot 11: 700.8m  96.2m
          3
=======================
slot 11: 655.7m  73.8m
          3
=======================
slot 11: 728.1m  75.3m
          3
=======================
slot 11: 604.9m  72.7m
          3
=======================
slot 11: 542.9m  21.7m
          3
=======================
slot 11: 621.4m  37.6m
          3
=======================
slot 11: 531.3m  22.0m
          3
=======================
slot 11: 572.0m  48.3m
          3
=======================
slot 11: 585.7m  27.5m
          3
=======================
slot 11: 561.1m  39.4m
          3
=======================
slot 11: 549.4m  41.3m
          3
=======================
slot 11: 475.3m  40.6m
          3
=======================
slot 11: 524.1m  59.0m
          3
=======================
slot 11: 556.6m  41.2m
          3
=======================
slot 11: 594.2m  47.3m
          3
=======================
slot 11: 565.1m  55.2m
          3
=======================
slot 11: 635.1m  65.5m
          3
=======================
slot 11: 548.0m  47.6m

 

 Comparing Link speed vs Link Utilization:

==========================

 

Info: 11/3 is 8gbps port.

 

hardly the highest value is 803.4m  or 1.2g  as total value of TX and RX. This says per second the through put is around 1gbps for this port.

 

But this is 8gbps port right? so in this case why its alerting as congestion on this port?

 

Please help me to get this understand.

Valued Contributor
Posts: 555
Registered: ‎03-20-2011

Re: Congestion Bottleneck -- No frame disc

[ Edited ]

First of all, portperfshow lists the values in kilo/mega BYTES per second, not it bits per second.

8 gigabits link is full duplex "8 gigabits TX" plus "8 gigabits RX". In your case, the TX line is clearly maxed out, reaching the ceiling of 800 megabytes per second.

What to do with that? It heavily depends on other details. From "relax, everything is fine" - e.g. if this is only due to backup activity and the job fits the backup window and other SLAs. Up to "go and fix it as fast as you can".

What is definitely wrong anyway, is that such overload only happens on one link in your SAN. Because normally multipathing should balance traffic across many (at least two) links. Or maybe you just didn't mention other overloaded links?

Contributor
Posts: 21
Registered: ‎05-26-2010

Re: Congestion Bottleneck -- No frame disc

Agree with Alexey.stepanov's response.  Portperfshow measures in BYTEs.  Brocade should comply with the known convention that smal k,m, g implies BITS and uppercase K, M, G implies BYTES.  Anyway, the TX reading show the 8Gb b/w is being achieve hence the bottleneck congestion alerts.

Join the Community

Get quick and easy access to valuable resource designed to help you manage your Brocade Network.

vADC is now Pulse Secure
Download FREE NVMe eBook