Managing IDS in IBM HACMP environment
Robert asked how to run IDS 10 on AIX 5.3 under HACMP: can one instance be shared between cluster nodes, and are there special considerations for the shared logical volumes holding dbspaces? Responses: with HACMP the binaries and chunks should both sit on the shared disks, with only one node active at a time — installing $INFORMIXDIR on shared disk works fine, though you must ensure files like /var/adm/ifx_loader_domain exist on both nodes. Others warned there's no ready-made IDS/HACMP agent (you script your own checks), noted split-brain risks, and suggested HDR (or SDS in IDS 11) as a simpler, faster-failover alternative; Robert planned HDR for his DR site.
Auto-generated by DrWatson from the posts below — may be imperfect; read the full thread.
About to install and setup IBS 10 on IBM p550/ AIX 5.3 HACMP environmnet. Are there any documents /guidelines for managing IDS in such environment? My concerns are (1) Can one IDS instance be shared between both nodes in the cluster or will IDS have to be installed and maintained individually on each node in the cluster? (2)Are there special considerations for management of the logical volumes on which IDS dbspaces reside? Hope to get some guidance on this. Regards, Robert
If you are planning to only failover IDS using the HACMP features, then the software as well as the data should reside on the shared disks. However, the instance will be actively running on only one machine at a time. Seriously consider using IDS's HDR features, which are included with all versions of IDS. In this way there will be an active IDS server instance on both machines and failover will be nearly instantaneous. This will require each machine to store its own copy of the data, with the concommittent additional disk costs, but as long as the secondary is only accessed when the primary is down, there is no additional Informix license or maintenance costs. IDS HDR secondary takeover tends to be significantly faster and more reliable than HACMP takeover and you get the added benefit of a fully redundant data store in addition to the redundancy of your RAID10 configuration on each machine. (Please tell me you are planning to use RAID1 or RAID10!) Art S. Kagel Oninit On Thu, May 8, 2008 at 11:52 AM, ROBERT ROMERO <rromero@tstt.net.tt> wrote: > About to install and setup IBS 10 on IBM p550/ AIX 5.3 HACMP environmnet. > > Are there any documents /guidelines for managing IDS in such environment? > > My concerns are > (1) Can one IDS instance be shared between both nodes in the cluster or > will > IDS have to be installed and maintained individually on each node in the > cluster? > > (2)Are there special considerations for management of the logical volumes > on > > which IDS dbspaces reside? > > Hope to get some guidance on this. > > Regards, > Robert > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > >
Just adding to Arts comments, HACMP and IDS work well. How long it takes so switch really depends on fast the VIP moves during the failover, and how long the instance takes to start. It is not as fast HDR but if you have external apps restarting, web servers etc then the failover time is much the same Cheers Paul -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of Art Kagel Sent: Thursday, May 08, 2008 11:05 AM To: ids@iiug.org Subject: Re: Managing IDS in IBM HACMP environment [12047] If you are planning to only failover IDS using the HACMP features, then the software as well as the data should reside on the shared disks. However, the instance will be actively running on only one machine at a time. Seriously consider using IDS's HDR features, which are included with all versions of IDS. In this way there will be an active IDS server instance on both machines and failover will be nearly instantaneous. This will require each machine to store its own copy of the data, with the concommittent additional disk costs, but as long as the secondary is only accessed when the primary is down, there is no additional Informix license or maintenance costs. IDS HDR secondary takeover tends to be significantly faster and more reliable than HACMP takeover and you get the added benefit of a fully redundant data store in addition to the redundancy of your RAID10 configuration on each machine. (Please tell me you are planning to use RAID1 or RAID10!) Art S. Kagel Oninit On Thu, May 8, 2008 at 11:52 AM, ROBERT ROMERO <rromero@tstt.net.tt> wrote: > About to install and setup IBS 10 on IBM p550/ AIX 5.3 HACMP environmnet. > > Are there any documents /guidelines for managing IDS in such environment? > > My concerns are > (1) Can one IDS instance be shared between both nodes in the cluster or > will > IDS have to be installed and maintained individually on each node in the > cluster? > > (2)Are there special considerations for management of the logical volumes > on > > which IDS dbspaces reside? > > Hope to get some guidance on this. > > Regards, > Robert > > > > **************************************************************************** *** > Forum Note: Use "Reply" to post a response in the discussion forum. > > **************************************************************************** *** Forum Note: Use "Reply" to post a response in the discussion forum.
Hi, - shared disks for the dbspaces (chunks) would be the easiest option. - shared disks for the installation of IDS binary may not be an option or may be fairly tricky or may be easy ... depends on the configuration of your cluster and your IDS instance(s). - the main problem would be the control of the IDS instance(s) during failover. I'm not aware that there is any ready-to-use software supporting IDS in a HACMP environment. (I may be wrong on this, though.) - in general the main problem with cluster failover is the "split brain" situation. Each node thinks it needs to run because the other seems to be dead, but this is not the case. Then you have to nodes running at the same time. Especially if they access the same shared disk, it is a disaster. - this situation normally is avoided by the STONITH method (Shoot The Other Node In The Head), to be sure that only one node is running. For all of this, some software is needed to control (or shoot) IDS ... The whole thing is anything but trivial! It can't be explained in detail in an e-mail (or two). Art Kagel is right in that HDR is (much) easier to setup and manage. This of course works with/for IDS only. HDR can't control other applications/services as it is built-in (IDS). Regards, Martin -- Martin Fuerderer IBM Informix Development Munich, Germany Information Management IBM Deutschland Entwicklung GmbH Chairman of the Supervisory Board: Martin Jetter Board of Management: Herbert Kircher Corporate Seat: Boeblingen, Germany Reg.-Gericht: Amtsgericht Stuttgart, HRB 243294 ids-bounces@iiug.org wrote on 08.05.2008 17:52:23: > About to install and setup IBS 10 on IBM p550/ AIX 5.3 HACMP environmnet. > > Are there any documents /guidelines for managing IDS in such environment? > > My concerns are > (1) Can one IDS instance be shared between both nodes in the cluster or will > IDS have to be installed and maintained individually on each node in the > cluster? > > (2)Are there special considerations for management of the logical volumes on > > which IDS dbspaces reside? > > Hope to get some guidance on this. > > Regards, > Robert > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum.
Putting the Informix files on the shared disk makes for easier maintenance
as you only have one area to maintain, it also gets a round the 'split
brain' situation as there can only be one master for the shared disks.
I'm not aware of any Informix HACMP software either, you can do trivial
tests like is oninit running and how many but this is a long way from ideal.
But historically we've always scripted our own tests and forced a failover
if they detect a real error - I can't remember at time when a failover was
triggered by IDS, if the engine crashes you can just restart it without
failing over - in fact you don't want to fail over
Cheers
Paul
-----Original Message-----
- shared disks for the dbspaces (chunks) would be the easiest option.
- shared disks for the installation of IDS binary may not be an option
or may be fairly tricky or may be easy ... depends on the configuration
of your cluster and your IDS instance(s).
- the main problem would be the control of the IDS instance(s) during
failover. I'm not aware that there is any ready-to-use software
supporting IDS in a HACMP environment.
(I may be wrong on this, though.)
- in general the main problem with cluster failover is the
"split brain" situation. Each node thinks it needs to
run because the other seems to be dead, but this
is not the case. Then you have to nodes running at
the same time. Especially if they access the same
shared disk, it is a disaster.
- this situation normally is avoided by the STONITH
method (Shoot The Other Node In The Head), to be
sure that only one node is running.
For all of this, some software is needed to control
(or shoot) IDS ...
The whole thing is anything but trivial!
It can't be explained in detail in an e-mail (or two).
Art Kagel is right in that HDR is (much) easier to setup and
manage. This of course works with/for IDS only.
HDR can't control other applications/services as
it is built-in (IDS).
Regards,
Martin
--
Martin Fuerderer
IBM Informix Development Munich, Germany
Information Management
IBM Deutschland Entwicklung GmbH
Chairman of the Supervisory Board: Martin Jetter
Board of Management: Herbert Kircher
Corporate Seat: Boeblingen, Germany
Reg.-Gericht: Amtsgericht Stuttgart, HRB 243294
ids-bounces@iiug.org wrote on 08.05.2008 17:52:23:
> About to install and setup IBS 10 on IBM p550/ AIX 5.3 HACMP
environmnet.
>
> Are there any documents /guidelines for managing IDS in such
environment?
>
> My concerns are
> (1) Can one IDS instance be shared between both nodes in the cluster or
will
> IDS have to be installed and maintained individually on each node in the
> cluster?
>
> (2)Are there special considerations for management of the logical
volumes on
>
> which IDS dbspaces reside?
>
> Hope to get some guidance on this.
>
> Regards,
> Robert
>
>
****************************************************************************
***
> Forum Note: Use "Reply" to post a response in the discussion forum.
****************************************************************************
***
Forum Note: Use "Reply" to post a response in the discussion forum.
If I'm using HDR I'd have all of the middleware running on both servers ready to serve clients instantly. No startup time, no failover delay. Art S. Kagel Oninit On Thu, May 8, 2008 at 12:56 PM, Paul Watson <paul@oninit.com> wrote: > Just adding to Arts comments, HACMP and IDS work well. How long it takes > so switch really depends on fast the VIP moves during the failover, and how > long the instance takes to start. It is not as fast HDR but if you have > external apps restarting, web servers etc then the failover time is much > the > same > > Cheers > Paul > > -----Original Message----- > From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of Art > Kagel > Sent: Thursday, May 08, 2008 11:05 AM > To: ids@iiug.org > Subject: Re: Managing IDS in IBM HACMP environment [12047] > > If you are planning to only failover IDS using the HACMP features, then the > software as well as the data should reside on the shared disks. However, > the instance will be actively running on only one machine at a time. > > Seriously consider using IDS's HDR features, which are included with all > versions of IDS. In this way there will be an active IDS server instance on > both machines and failover will be nearly instantaneous. This will require > each machine to store its own copy of the data, with the concommittent > additional disk costs, but as long as the secondary is only accessed when > the primary is down, there is no additional Informix license or maintenance > costs. IDS HDR secondary takeover tends to be significantly faster and more > reliable than HACMP takeover and you get the added benefit of a fully > redundant data store in addition to the redundancy of your RAID10 > configuration on each machine. (Please tell me you are planning to use > RAID1 or RAID10!) > > Art S. Kagel > Oninit > > On Thu, May 8, 2008 at 11:52 AM, ROBERT ROMERO <rromero@tstt.net.tt> > wrote: > > > About to install and setup IBS 10 on IBM p550/ AIX 5.3 HACMP environmnet. > > > > Are there any documents /guidelines for managing IDS in such environment? > > > > My concerns are > > (1) Can one IDS instance be shared between both nodes in the cluster or > > will > > IDS have to be installed and maintained individually on each node in the > > cluster? > > > > (2)Are there special considerations for management of the logical volumes > > on > > > > which IDS dbspaces reside? > > > > Hope to get some guidance on this. > > > > Regards, > > Robert > > > > > > > > > > **************************************************************************** > *** > > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > > > > **************************************************************************** > *** > Forum Note: Use "Reply" to post a response in the discussion forum. > > > > ******************************************************************************* > Forum Note: Use "Reply" to post a response in the discussion forum. > >
Appreciate the comments. We are planning to use HDR but this will be to our Disaster Recovery Site. The use of HACMP at the Primary site is mainly to have high availability in case of any hardware failure and for system maintenance.The cluster is already setup and shared logical volume are being created for the dbspaces but I am uncertain about whether the IDS installation can reside on shared volume. Robert.
you can install IDS on the shared disks - works fine Cheers Paul -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of ROBERT ROMERO Sent: Thursday, May 08, 2008 3:59 PM To: ids@iiug.org Subject: Re: RE: Managing IDS in IBM HACMP environment [12053] Appreciate the comments. We are planning to use HDR but this will be to our Disaster Recovery Site. The use of HACMP at the Primary site is mainly to have high availability in case of any hardware failure and for system maintenance.The cluster is already setup and shared logical volume are being created for the dbspaces but I am uncertain about whether the IDS installation can reside on shared volume. Robert. **************************************************************************** *** Forum Note: Use "Reply" to post a response in the discussion forum.
I believe I ran across a gotcha with this. You need to make sure that the ifx_loader_domain file exists on both boxes. This file gets created on the install. On our AIX systems it is created in /var/adm. You can grep the steps to create it from the $INFORMIXDIR/etc/install. Just make sure you test and you should find any other gotchas. David -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of Paul Watson Sent: Thursday, May 08, 2008 4:08 PM To: ids@iiug.org Subject: RE: RE: Managing IDS in IBM HACMP environment [12054] you can install IDS on the shared disks - works fine Cheers Paul -----Original Message----- From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of ROBERT ROMERO Sent: Thursday, May 08, 2008 3:59 PM To: ids@iiug.org Subject: Re: RE: Managing IDS in IBM HACMP environment [12053] Appreciate the comments. We are planning to use HDR but this will be to our Disaster Recovery Site. The use of HACMP at the Primary site is mainly to have high availability in case of any hardware failure and for system maintenance.The cluster is already setup and shared logical volume are being created for the dbspaces but I am uncertain about whether the IDS installation can reside on shared volume. Robert. ************************************************************************ **** *** Forum Note: Use "Reply" to post a response in the discussion forum. ************************************************************************ ******* Forum Note: Use "Reply" to post a response in the discussion forum.
> -----Original Message----- > From: ids-bounces@iiug.org [mailto:ids-bounces@iiug.org] On Behalf Of > ROBERT ROMERO > Sent: Thursday, May 08, 2008 1:59 PM > To: ids@iiug.org > Subject: Re: Managing IDS in IBM HACMP environment [12053] > > Appreciate the comments. > > We are planning to use HDR but this will be to our Disaster Recovery > Site. The > use of HACMP at the Primary site is mainly to have high availability in > case > of any hardware failure and for system maintenance.The cluster is > already > setup and shared logical volume are being created for the dbspaces but > I am > uncertain about whether the IDS installation can reside on shared > volume. > > Robert. If you can go to IDS 11x, then it sounds like a job for SDS (shared disk secondary). HTH, Paul M.