UNSOLVED

bhansen

updated

11 years ago

B

bhansen

54 Posts

0

3898

June 22nd, 2009 08:00

HBA Login to FA Port

I have an AIX host, that has two HBA cards, Emulex installed, part of a HA cluster. It is zoned and had disk space masked from two different DMX frames, once a DMX3, one a DMX4. On one FA port on the DMX4, the HBA is no longer logged into it. This same HBA is how ever logged into the FA port on the DMX3. Plus the 2nd node of the cluster is zoned and mapped to the same FA ports on the DMX3 & DMX4, and is logged into each FA port. Is there a command I can have the AIX admin run to force the HBA to log back into the one FA port?
  • Allen Ward

    6 Operator

    2124 Posts

    1521

    0

    Posted June 22nd, 2009 08:00

    In general, we have the AIX admins run a cfgmgr command to rescan the busses for connections. If the HBA is not logged in to any switch ports or array FAs this usually resolves it for us.

    However, I am not a UNIX guru and I don't know if there is a better way to handle this. I just know what works for me. For all I know I may be asking them to do this and they could be laughing their heads off and doing something entirely different (and correct). I doubt it, but you never know :-)
  • dynamox

    11 Legend

    20419 Posts

    87439 Points

    1521

    0

    Posted June 23rd, 2009 05:00

    cfgmgr should do it ..i am part-time aix admin :) and whenever i zone an AIX box to another array, i have to run cfgmgr to force to login.
  • MikeMac1

    2 Intern

    292 Posts

    1521

    0

    Posted June 24th, 2009 05:00

    On the DMX4 where the HBA is not logged in, do you have other HBA's currently logged in? If you don't, you may have an issue between the switch and the FA.
  • bhansen

    54 Posts

    1521

    0

    Posted June 24th, 2009 07:00

    AIX admin run CFGMGR, now we are logged back int the FA port, but the host still reports that the path is down. We do have this HBA active to a 2nd array, as well the FA pport has the other half of this AIX cluster functioning fine. Seems these two items eliminate the fiber, HBA, switch ports and FA ports, at least on the surface.

    His response from IBM are all that there is a failed component in the SAN, not sure I can agree with this knowing all the other items are working, but not going to rule anything completely out.
  • MikeMac1

    2 Intern

    292 Posts

    1521

    0

    Posted June 24th, 2009 07:00

    Brian, I'm sending you an e-mail to gather some information so I can take a look for you.
  • Allen Ward

    6 Operator

    2124 Posts

    1528

    0

    Posted June 24th, 2009 08:00

    I'm going to defer to one of the other participants who are more familiar with AIX on how to best interpret the output. I know what I would do if you were running Powerpath, but it sounds like you are not.
  • Allen Ward

    6 Operator

    2124 Posts

    1521

    0

    Posted June 24th, 2009 08:00

    LOL - Sounds like a typical IBM/Server Support answer. Everyone used to blame things on Network & Server, but now that Storage is a seperate discipline we seem to get blamed for everything first :-)

    Have you checked to see if AIX created a new path for this connection and is just reporting the old one as dead? Are you using PowerPath, MPIO, or some other multipathing?
  • bhansen

    54 Posts

    1521

    0

    Posted June 24th, 2009 08:00

    Do you know the process to follow to check to see if the host created a new path. I do not have a high level of confidence the particular AIX admin I am dealing with will know.
  • Allen Ward

    6 Operator

    2124 Posts

    590

    0

    Posted June 24th, 2009 08:00

    Well, that is different then. Run "powermt display dev=all" and check the paths on any of the DMX devices.

    You should be able to see if there is a new "alive" path which has replaced an "dead" path. If you still don't see your new path, go back and double check the masking for this one HBA through this FA. I'm going to assume that the mapping is correct since it sounds like you can see the devices properly through the other host in the cluster. If my assumption is wrong, then double-check your mapping too.
  • bhansen

    54 Posts

    1053

    0

    Posted June 24th, 2009 08:00

    We are running power path