Re: HACMP question

1) Without GPFS, active/active concurrency I use:

cluster.es.cfs.rte
cluster.es.client.lib
cluster.es.client.rte
cluster.es.client.utils
cluster.es.client.wsm
cluster.es.cspoc.cmds
cluster.es.cspoc.dsh
cluster.es.cspoc.rte
cluster.es.plugins.dhcp
cluster.es.plugins.dns
cluster.es.plugins.printserver
cluster.es.server.cfgast
cluster.es.server.diag
cluster.es.server.events
cluster.es.server.rte
cluster.es.server.testtool
cluster.es.server.utils
cluster.msg.en_US.cspoc
cluster.msg.en_US.es.client
cluster.msg.en_US.es.server

2) Not a GPFS guy

3) Most boxes have multiple subnets barring webservers. For each VLAN
your should have a minimum of 3 IP's across two subnets (total 3).
Assuming aliasing whith is the v5+ default.
example:
10.10.30.100 Boot
10.11.30.100 Persistent (used for cluster communication on mine, but
not necessarily)
10.11.30.101 Service
Boot and service must be on the same VLAN but seperate subnets.
Assuming you create a special even non-routed boot subnet like most do,
you probably want the service and persistent on the same routable
subnet.
The role of the boot is to verify the service subnet's functionality in
the even of IP network fubar.

4) You can HB across TTY (not recommended), IP and SAN. With SAN
allocate a single LUN and dont use it for storage. I pop mine in a
concurrent VG that never varries on. TTY isn't recommend simply because
it doesn't lend itself to virtualization at all. You have a fine 32bit
card and a 50 ft cable as thick as my thumb running between two bricks
of approximately the same size as 100 iTouch devices, each. Not too
onerous with dual 595's. An enormous PITA with dual full racks of
550's. To each his own.

5) I think you used standby when you should have said service

We have three ip addresses on both nodes:
Boot ip which is there when the system boots, and before
cluster services is started.
Persistent ip which is always there regardless of the
cluster services state, for admin management of the node.
Standby ip which belongs to a resource group to be moved
from aixoradb1(home node), to aixoradb2 in the event of a failover.

John T Mills

________________________________

From: IBM AIX Discussion List [mailto:aix-l@xxxxxxxxxxxxx] On Behalf Of
Justin Bleistein
Sent: Thursday, February 26, 2009 7:58 AM
To: aix-l@xxxxxxxxxxxxx
Subject: HACMP question


I have a few basic questions for you HACMP guys.

1.) When you guys are installing HACMP at a client site like from CD, do
you guys just install all filesets
with the name cluster.*? The install guide lists all of the filesets
which come with AIX, but deson't tell me what I should install exactly.
There seems to be alot of unnecessary filesets on the media for just
like a base HACMP install. Could you list the filesets you install?

2.) GPFS. This is from what I understand. You have a shared SAN disk
with like an Oracle database, or something you are making highly
available with HACMP. You have the vg which is on the disk imported, on
both nodes, for fast disk takeover. Now, GPFS takes it a step forward,
and allows you to not only have the disk imported in active-passive mode
on both nodes, but allows you to mount that resource group's filesystems
on both nodes simultaneously. GPFS coordinates the read/write activity
to the disk by both nodes for consistency.

3.) I'm a little confused on the network interfaces. I know there is a
boot interface, which I believe is also known as the base (Is that
correct)? Some cluster configs I've been reading about have another
boot, what's up with that? There is a persistent which I'm pretty clear
on, and a service ip which I totally understand. The boot ip is kind of
throwing me for a loop though. Could you explain that? What is the boot
ip, and what is it used for? The only thing I could put together is that
the boot ip is the ip which is on the interface before the HACMP cluster
services daemons are started on node, but there has to be more to it's
role. What is its role in the cluster? What is the state of that ip
address(boot), when the cluster services daemons are started?

4.) Where does the heartbeat/keep alive packets travel through? What
interface?

5.) Could you provide a real basic diagram of a typical HACMP cluster in
AIX? I attempted to start the following text diagram. If you could fill
in the rest, or correct any portion of it where I might be wrong, I'd
appreciate it:

Two AIX nodes with the AIX bos installed locally on internal disks -
rootvg.
There is a disk presented/zoned to the wwns of both nodes via the SAN.
Both nodes see the same disk.
The HACMP software is installed locally on both nodes
Both nodes belong to a standby cluster I'll call "ora_aix_ha_clus1"
The two node names which are part of the cluster are named after their
hostnames to keep things sane - aixoradb1, aixoradb2.
The Oracle database is installed on the SAN disk, oracle home, and the
Oracle database files, dbfs, control files, redo logs, archive logs,
etc., reside on the shared disk on the SAN.
We have three ip addresses on both nodes:
Boot ip which is there when the system boots, and before
cluster services is started.
Persistent ip which is always there regardless of the cluster
services state, for admin management of the node.
Standby ip which belongs to a resource group to be moved from
aixoradb1(home node), to aixoradb2 in the event of a failover.
The shared vg will be imported to both nodes. It will be varied on to
aixoradb1 in active mode with filesystems mounted, and in passive mode
on aixoradb2 with no filesystems mounted.

In the event of a failover the following will happen:

The vg currently in passive mode on aixoradb2 will be varied on to
active mode.
The filesystems will all be mounted making the Oracle files visible to
aixoradb2.
The application server startup script will be started which will
re-start the Oracle database on the shared disk on aixoradb2 which will
perform instance recovery, and bring that Oracle database up again.
The service ip address via alias replacement, since that seems to be
recommended over replacement, will be configured onto an interface on
aixoradb2.
Now the Oracle client applications who were previously connected to the
database via aixoradb1, will be knocked off of their connection briefly,
but once the resource group is completely moved to aixoradb1,
then when the clients reconnect their application again it will
reconnect to the database again. It will be transparent to them however
that the ip address actually moved, and they are talking to another
physical
server which is hosting the same database.

This was obviously a very simple scenario. I just wanted to ensure that
I understood it from a very high level.


Please answer, and confirm when you get a moment.

Thanks in advance.

Justin Richard Bleistein - AIX/TSM/Oracle Database Administrator

Relevant Pages

  • SUMMARY: Restoring cluster from tape
    ... However, after about a week of effort on my part, I still didn't have a bootable cluster and decided instead to concentrate on migrating all our users, data and applications to the Operton system that we had already bought to replace the alphas. ... A RAIDset on the HSG died recently, which contained the cluster_root, cluster_usr, cluster_var, root1_domain and root2_domain AdvFS domains AND the quorum disk. ... Of course the WWIDs of the new partitions are not the same as their original equivalents, so I have had to use wwidmgr at the SRM prompt to enable the new boot devices. ... halted CPU 0 ...
    (Tru64-UNIX-Managers)
  • SUMMARY: restore a crashed test cluster
    ... I have a 5.1B cluster with dupatch 3. ... the quorum disk. ... and restore the following ... I then go to the boot prompt again, and try to start the first member by ...
    (Tru64-UNIX-Managers)
  • Re: BIOS doesnt find boot record
    ... Did you set your BIOS to boot from SCSI disk? ... Reserved Sectors: ... MFT Start Cluster: ...
    (microsoft.public.windowsxp.help_and_support)
  • Re: how to break cluster
    ... and the machine will be out of the cluster. ... There are big caveats to this: If node B relies on access to drives ... Also note that a node may rely on another node's system disk to boot. ...
    (comp.os.vms)
  • Re: HACMP question
    ... The vg currently in passive mode on aixoradb2 will be varied on to ... Depends on how you setup the cluster and what options you ... So the alias will bind to 1 during HACMP startup. ... Shared disk. ...
    (AIX-L)