Blades not staying logged in via Virtual Connect FC

[Magnus Andersen]

I have the following configuration.

• c7000 chassis
• 6 BL460 blades with mezzanine card active
• 1 VC Flex-10 Enet module, that works great (Firmware 3.15)
• 1 VC 8GB 20-port FC module (Firmware 1.41)
• 1 Brocade 300 FC switch (Fabric OS v6.4.1b)
• 1 P2000 G3 FC SAN

I have the 4 uplink ports connected to the ports on the Brocade 300 switch and they are working and logged in.  The HBAs in the blades are mostly in NOT-LOGGED-IN state.  If I power cycle the VC module the HBAs sometimes change their status, but even if all of the log in (which has happened), 30 minutes later half of the HBAs are back in the NOT-LOGGED-IN state.  There isn't 1 blade or port that always have issues.  It is totally random to me.  For the HBAs that are in the NOT-LOGGED-IN state I can see a FE login on the switch, but not an FF login.  The P2000 has no issues staying connected to the Brocade 300 switch.  I have several LUNs presented to the servers that periodically revert to the NOT-LOGGED-IN state.

Any ideas how to get the blade HBAs to log in and to stay logged in?

Thanks,
Magnus 

[Psychonaut]

Under your SAN Fabrics - what is your Login Re-Distribution set to? 

[Magnus Andersen]

It is set to Manual.  I don't see a place to change it either.

[Psychonaut]

Actually disregard that - I was thinking in FlexFabric terms.

[Johan Guldmyr]

Sometimes fixing the port speed on the SAN-ports help, perhaps at 4G would be worth a shot in your case.

Consider it being on the OS side, drivers or HBA firmware perhaps?

Isn't it so that the blade won't show up as LOGGED IN until the OS has booted?

[JKytsi]

And Your Brocade ports have NPIV enabled ?

[Magnus Andersen · Accepted solution]

Thanks for all ideas.  I ended up downgrading the firmware to 2.31 on the Flex-10 Enet module and to firmware 1.40 on the 20-port FC module.  Now works without issues.

[JKytsi]

Something must have gone wrong in previuous update. Version 2.31 is really old, try to update to latest 3.3x version.

[JD-WFH]

Same issue here. We had an older Brocade switch and had to change the Fill Word parameter. It didn't take on a vSphere host, but we're guessing a reboot will do the trick.