Re: Migrating from one server to another

From: Robert Miller (rmiller_at_SMUD.ORG)
Date: 06/28/04

  • Next message: Ayman Rayyan: "Mount jfs2 on sun"
    Date:         Mon, 28 Jun 2004 08:10:06 -0700
    To: aix-l@Princeton.EDU
    
    

    Oh heck no. M$ clients and HA boxen on the same subnet? I cringe to
    even consider that...

    No, this is for the clients on the other side of a router - I believe
    that the router just passes the ARP request on, but the client rejects
    it (or at least that's how it sounded from what the instructors were
    saying). I don't pretend to actually understand it (until I have time
    to research it better), but two different instructors have said the same
    thing, so there must be a grain of truth in there somewhere ;)

    --rm

    -----Original Message-----
    From: IBM AIX Discussion List [mailto:aix-l@Princeton.EDU]On Behalf Of
    Holger.VanKoll@SWISSCOM.COM
    Sent: Monday, June 28, 2004 8:03 AM
    To: aix-l@Princeton.EDU
    Subject: Re: Migrating from one server to another

    so you have windows-clients in the same subnet as the servers running
    hacmp?!?

    if you do so (and cannot move them out), you could add them to the
    pinglist

    check with a simple ping before to see if they will refresh the
    arp-cache

    -----Original Message-----
    From: IBM AIX Discussion List [mailto:aix-l@Princeton.EDU] On Behalf Of
    Robert Miller
    Sent: Monday, June 28, 2004 4:49 PM
    To: aix-l@Princeton.EDU
    Subject: Re: Migrating from one server to another

    Fresh from a set of HACMP classes, I have a bit more to offer on this,
    at least according to the instructors (I haven't dug into the networking
    stuff to figure out where gratuitous ARP is yet...).

    According to the instructors, the problem lies with the *windows*
    clients (imagine that...), and that they are the ones that will not
    accept the gratuitous ARP requests to refresh their cache. Not sure yet
    where to look for this setting, perhaps the M$ support site will shed
    some more light on the subject...

    --rm

    -----Original Message-----
    From: IBM AIX Discussion List [mailto:aix-l@Princeton.EDU]On Behalf Of
    Coonfield, Ben
    Sent: Friday, June 18, 2004 12:39 PM
    To: aix-l@Princeton.EDU
    Subject: Re: Migrating from one server to another

    I'm using the same function. The latest versions of HACMP not only
    tolerate but use IP aliases, that is, HACMP automatically moves an IP
    alias from one box to another automatically.

    The switch issue is because using this function requires a function
    called "Gratuitous ARP". Basically, this means that when you add an IP
    alias, the target broadcasts an ARP reply packet to tell the routers
    where the IP address is now located. It is exactly like a regular ARP
    reply, it is called gratuitous because there was no arp request, just
    the reply. I believe that most network hardware handles this with no
    problem, but some switches might need to be configured to ensure that
    they will permit this.

    -----Original Message-----
    From: IBM AIX Discussion List [mailto:aix-l@Princeton.EDU]On Behalf Of
    Green, Simon
    Sent: Friday, June 18, 2004 2:08 PM
    To: aix-l@Princeton.EDU
    Subject: Re: Migrating from one server to another

    We had some problems at the start with the network switches, because
    they weren't updating their arp cache when the alias got moved to a
    different box: took thirty minutes to refresh, or taking the interface
    down and up again. That was sorted out by one of my colleagues, but I
    forget how. Since I'm going home in a minute, I can't be bothered to
    look it up, either. :-)

    We've also had a problem on some servers where there are lots of aliases
    - >10? - where one of the aliases seems to become the "primary" address
    - top of the list in netstat -i. When you unconfigure that alias to
    move it somewhere, it takes the default gateway with it and also messes
    up the routing table in some as-yet vague and undetermined way! (I
    think that it retain an entry for that interface with the alias address,
    instead of having the correct non-alias address.) I've only seen this
    once for sure, but I think it's actually happened twice. We haven't
    been able to reproduce the problem, though.

    We had a problem once when we added an application alias that was in a
    different subnet to the primary. Subsequently, connections to other
    nodes in that second subnet failed, because IP thought they were local,
    and didn't send them to the gateway. That's not really a problem with
    the concept, though, just an administrative error.

    There may be some issues with older versions of HACMP, but I think
    current ones should cope with it.

    rsh, ftp etc work fine to the aliases. We don't do NFS mounts using the
    aliases, though. (I think that could cause some confusion with exports,
    although I don't think it would be difficult to sort out. But we intend
    the aliases to be used by *users*, not for actually configuring stuff on
    the servers themselves.)

    Aliases are added through a script we set up, with a centrally
    maintained list of aliases. We do not have them in the ODM. That makes
    it easier to move them around without having to remove the ODM
    entry: just change the master config file and push it out with rdist.

    We're very happy with the way it all works.

    --
    Simon Green
    Altria ITSC Europe Ltd
    AIX-L Archive at https://new-lists.princeton.edu/listserv/aix-l.html
    New to AIX? http://publib-b.boulder.ibm.com/redbooks.nsf/portals/UNIX
    N.B. Unsolicited email from vendors will not be appreciated.
    Please post all follow-ups to the list.
    > -----Original Message-----
    > From: Bill Verzal [mailto:BVerzal@KOMATSUNA.COM]
    > Sent: 18 June 2004 19:44
    > To: aix-l@Princeton.EDU
    > Subject: Re: Migrating from one server to another
    >
    >
    > Nice concept.  I think I'll store that in the back of my head.  Have
    > you come across any issues with that procedure ?
    -----------------------------------------
    ________________________________________________________________________
    ____________
    ****CONFIDENTIALITY NOTICE****
    This e-mail is the property of The Methodist Hospital and/or its
    relevant affiliates and may contain confidential and privileged material
    for the sole use of the intended recipient(s). Any review, use,
    distribution or disclosure by others is strictly prohibited. If you are
    not the intended recipient(or authorized to receive for the recipient),
    please contact the sender and delete all copies of the message.  Thank
    you.
    

  • Next message: Ayman Rayyan: "Mount jfs2 on sun"

    Relevant Pages

    • Re: dhcprelay troubleshooting, where next?
      ... > It quite happily supplies addresses to everything on that subnet. ... The server has noted this in its dhcpd.leases file. ... > Listening on Socket/eth1 ... > plus lots of inaddr arp domain unreachable and PTR? ...
      (comp.os.linux.networking)
    • RE: arp
      ... YOu can try rebooting the server into Safe mode with Networking and if the ... compare the processes running vs the processes running in a ... >to broadcast allot of ARP for the whole subnet? ... It will bring the whole subnet down. ...
      (microsoft.public.win2000.security)
    • arp
      ... Is it normal for Windows 2000 server that is in NT domain as a member server ... to broadcast allot of ARP for the whole subnet? ... It will bring the whole subnet down. ...
      (microsoft.public.win2000.security)
    • Re: NLB Cluster - Ping fails or long time to reply from outside local subnet - SOLVED
      ... Once again, ARP is an RFC standard, if you are having to make static entries in unicast mode, then your network device is not in compliance. ... Windows Server 2008 Readiness Team ... I was feeling nervous about our teaming-capable adapter as I read it might be sending out heartbeats, so I disabled it AND configured the cluster on a separate DLink card in multicast mode. ... I thought that the litmus test was that the router functions fine when no NLB is installed, but when it is, things start going screwy. ...
      (microsoft.public.windows.server.clustering)
    • Re: NLB Cluster - Ping fails or long time to reply from outside local subnet - SOLVED
      ... ARP is defined by RFC. ... Windows Server 2008 Readiness Team ... The servers are a couple of switches away from the router so I would have thought that any duplicate MAC info. or similar would have been come irrelavent/concealed at the router. ... If you run the command 'wlbs query' and the node it is run on says it is converged with all of your nodes in the cluster, that basically says the NLB configuration is correct and the nodes are talking to each other. ...
      (microsoft.public.windows.server.clustering)