Comments (10)
  • Add a Comment
  • Edit
  • More Actions v
  • Quarantine this Entry

1 seb_ commented Permalink

Good blog! I still see that often. Just some additional remarks: <br /> - In fact mode 2 is according to the FC standard. But as HBA Vendors already had to cope with the wrong Brocade implementation, there are devices out there relying on the wrong way to do it. So as you already stated, mode 3 is the best compromise. <br /> - Although described to affect all connections, this setting is only important for Non-ISL ports. If you have for example a longdistance ISL, the fillword depends on the VC Link Init parameter. You will even get a warning if your longdistance config is contrary to the fillword setting, but it's meaningless. But of course, if you change the fillword modes on your switches anyway, there's no harm to have the ISLs on mode 3, too. <br /> - If you have no chance to update to the latest codes and have to stick with either mode 0 or 1, mode 0 is recommended for 8G ports. Dont worry if you see "Invalid transmission words" (bad_os) increasing in portstatsshow. The switch just complains about getting the wrong fill word here. That's not a problem unless you have portfencing enabled, which could cause the switch to automatically disable the port because of it. If you use 8G with fillword mode 0, don't use portfencing.

2 anthonyv commented Permalink

Thanks Seb! Excellent additional information.

3 RainerWinnen commented Permalink

Thx - great information - solved my Brocade <-> Storwize V7000 Problem

4 avandewerdt commented Permalink

Glad to hear!

5 minosi commented Permalink

A bit later but the topic seems to be alive still. :) <div>&nbsp;</div> So it should be reset to 3 for all F-ports - even those that generate a lot of er_bad_os when on 3 instead of 2 ?

6 avandewerdt commented Permalink

Frankly use the setting that gives stable running with no errors. <br /> Seb (the first comment) says to ignore bad_os errors, see his comment. <br />

7 minosi commented Permalink

Thank you Anthony for the quick reply.

There came a global initiative which considers anything not at 3 "dangerous", so we are asked to go from no-error setup of
0 <=4G devices
2 8G hosts (Qlogic)
3 8G subsystems
to all 8G Fs on 3.
But if bad_os can be fully ignored, then 3 looks like the (right) KeepItSimpleSilly way.

8 ErlendØyen commented Permalink

Hello <div>&nbsp;</div> Great info! <div>&nbsp;</div> Does this setting apply to uplink ports between FC switches as well? <div>&nbsp;</div> What about regular host with 8gb hba?

9 avandewerdt commented Permalink

It does not apply to hosts or uplink ports as far as I know, thought most people chose to set every port to '3'.

10 robert_ commented Permalink

I noticed on one port that was getting < .02 CRC errors that is was also getting a lot of the er_bad_os/Invalid ordered set. This got me interested to see what the other ports were doing and I noticed many of my 8 gig ports were reporting high invalid_ordered_set errors without having crc errors. Other than the high count, what side effects can this cause on the channel? I have seen some people say to ignore it and other say it can cause resets on the channel. I think I need to change my fillword from 0 to 3 but need to have a good reason for management approval. Running v7.2.1 on dcx-4s and 6.4.2b on 5470 switch. Thanks!