Fibre Channel (SAN)

Reply
New Contributor
Posts: 4
Registered: ‎03-04-2014

Server Latency bottleneck detected at slot 0 port 40 alert in Brocade 5300 SAN switch.

Hi,

 

We are gettings alerts message "Server Latency bottleneck detected at slot 0 port 40 alert in Brocade 5300 SAN switch".

 

Please help me in resolving the issue.

Occasional Contributor
Posts: 18
Registered: ‎05-26-2010

Re: Server Latency bottleneck detected at slot 0 port 40 alert in Brocade 5300 SAN switch.

[ Edited ]

Latency bottleneck means the attached device (server HBA) is not able to receive frames at the delivery rate of the 5300. This is also known as a "slow-drain" situation whereby the server's HBA is not able to keep up with feed rate of the 5300.
Many possible causes. I would suggest inspecting what the server is doing at timestamp of latency messages. Is a high demand application reading from storage? Is the HBA current in f/w and driver? Is the application and/or HBA queue depth set appropriately? It could also be the HBA itself. In any of the situation the 5300 port is not receiving R_Rdys back fast enough and causing a latency condition (aka back-pressure) from server to 5300. The problem is on the server end.

Valued Contributor
Posts: 551
Registered: ‎03-20-2011

Re: Server Latency bottleneck detected at slot 0 port 40 alert in Brocade 5300 SAN switch.

SERVER or SEVERE ???
New Contributor
Posts: 4
Registered: ‎03-04-2014

Re: Server Latency bottleneck detected at slot 0 port 40 alert in Brocade 5300 SAN switch.

 

Thanks for replying back.

 

Actually, the port is connected Bladecenter Switch 5480 ( ISL).

Valued Contributor
Posts: 551
Registered: ‎03-20-2011

Re: Server Latency bottleneck detected at slot 0 port 40 alert in Brocade 5300 SAN switch.

No, I mean, does your error read "server latency bottleneck ..." or "severe latency bottleneck ..."

Because I'm sure it should be "severe"

Do you see increase of any counters on that port #40 in portshow or in portstatsshow?

Now that you said this port actually cascades to the blade chassis, do you see any increasing errors in that blade switch?

I bet from my experience that you have some CRCs in there...
Occasional Contributor
Posts: 18
Registered: ‎05-26-2010

Re: Server Latency bottleneck detected at slot 0 port 40 alert in Brocade 5300 SAN switch.

Ok. If the FCSM in bladecenter does not run FOXS with bottleneck feature, then to find which blades are slow you could monitor the tim_crd_z counter vs. total frames transmitted on FCSM towards the blade HBAs. If the ratio of tim_crd_z counts (found in portstatsshow X) is high vs. frames tx count, then likely the server is slow.
However, beware the counters are 32bit and can wrap quickly. Clearing the FCSM counters and taking a reading within 5mins will ensure neither counter wrapped to ensure your calculations of tim_crd_z vs. frames tx is accurate.

Join the Community

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

Download FREE NVMe eBook