Clustering and Load balancing in OE RDBMS (RH Linux)

Posted by tharanga_herath on 19-Feb-2013 21:15

Hi,

We have a site that use Sonic ESB 8.5.1 and backend Database is Progress OE 10.2B with fathom replication.

All components are installed in linux boxes.
Database calls are making through the Sonic  DBservice and separate SQL broker is available to handle the load.
I got to know that OE database not support for clustering in Linux environment and We need to do scripting our Owen and I don’t have any Idea of how to do it and the stability of such a implementation.
Is anyone having experience on enabling database clustering on Linux OS and method of load balancing apart from redirecting Reporting Applications to DR server?

All Replies

Posted by Valeriy Bashkatov on 20-Feb-2013 00:28

Oh, that's a problem. For that year, I ask the PSC to made the Procluster for linux, but so far without success. That's the last time in Manchester I spoke to them about it and they promised they would do, but that's not clear when .... very bad I do not understand how they will work in the cloud, if they do not have an elementary clustering, not to mention the Real Application Cluster (RAC)

Posted by tharanga_herath on 20-Feb-2013 01:02

Agreed,  I also contact the progress support on this and come out the same answer. This is bit issue for us now when we dont have real loadbalancing and clustering in progress RDBMS. Where other DB vendors offers. I feel this is a right time that progress to address these issues. .

Posted by Jarek Kowalewski on 30-Sep-2014 18:55

I asked several times about failover cluster support in RHEL during EMEA PUG challenge BOF sessions but everytime they (Progress fellows) seem to be surprised and they look at me like I was on another planet. Four/five years ago I also send question about cluster to the PEGgers, but no luck to see any answer.

In 2010 I made some successful tests on AIX (HACMP) and Windows Server 2003 Enterprise (MSCS). I tried also with Red Hat Cluster Suite in the unsupported way and the results were quite good. OE database is not cluster aware. We used rc.d script to let RHCS to control database. The only big challenge was to write appropriate rc.d script with start|stop|monitor cases and exact return values.

After some tests the only implementation differences between supported & unsupported CMs I'd noticed is that procluster "exposes" database resources to the CM and do some changes in master block of db, so it is cluster aware.

I'm using killall -9 _mprosrv in rc.d script to stop resource, but it's not recommended. Process of relocating works fine. The worse scenario is when active node hangs, then second node cannot start db due to hostname entry in lock file.

In my opinion if the RHEL is certified for OE, the RH Cluster Suite should also be certified cluster manager for OE. In database environment the failover cluster is a basic method to increase reliability and safety and I'm wondering why it is so neglected by PSC.

Posted by Jarek Kowalewski on 30-Sep-2014 18:59

AFAIK you won't succeed in load balancing because OE allow active/passive type of clustering.

Posted by Vinko Buzak on 01-Oct-2014 01:00

You are correct,
 
OpenEdge currently only supports the following Failover Cluster Managers:
 
- Microsoft Windows Server 2008 R2 Failover Clusters (64-bit OpenEdge only)
- Microsoft Cluster Services MSCS V5.2 Windows Server 2003 R2 Enterprise Edition
- IBM HA Cluster (HACMP) 5.3
- IBM HA Cluster (HACMP) 5.4.1 with OS AIX 5.3
- IBM HA Cluster (HACMP) 6.1 with OS AIX 6.1
- Sun Solaris Sun Cluster 2.2, 3.0, 3.1, & 3.2
- HP-UX Service Guard 11i (PA-RISC)
- HP-UX Service Guard 11i (Itanium2)
 
Therefore, you may need to talk with your Progress account manager and ask when will Progress certify failover cluster support in Linux RedHat, as that is “the only” Unix-flavour missing this certification :-)
 
[collapse]
From: jkowalewski [mailto:bounce-jkowalewski@community.progress.com]
Sent: Wednesday, 1 October 2014 12:56 p.m.
To: TU.OE.RDBMS@community.progress.com
Subject: RE: [Technical Users - OE RDBMS] Clustering and Load balancing in OE RDBMS (RH Linux)
 
Reply by jkowalewski

I asked several times about failover cluster support in RHEL during EMEA PUG challenge BOF sessions but everytime they (Progress fellows) seem to be surprised and they look at me like I was on another planet. Four/five years ago I also send question about cluster to the PEGgers, but no luck to see any answer.

In 2010 I made some successful tests on AIX (HACMP) and Windows Server 2003 Enterprise (MSCS). I tried also with Red Hat Cluster Suite in the unsupported way and the results were quite good. OE database is not cluster aware. We used rc.d script to let RHCS to control database. The only big challenge was to write appropriate rc.d script with start|stop|monitor cases and exact return values.

After some tests the only implementation differences between supported & unsupported CMs I'd noticed is that procluster "exposes" database resources to the CM and do some changes in master block of db, so it is cluster aware.

I'm using killall -9 _mprosrv in rc.d script to stop resource, but it's not recommended. Process of relocating works fine. The worse scenario is when active node hangs, then second node cannot start db due to hostname entry in lock file.

In my opinion if the RHEL is certified for OE, the RH Cluster Suite should also be certified cluster manager for OE. In database environment the failover cluster is a basic method to increase reliability and safety and I'm wondering why it is so neglected by PSC.

Stop receiving emails on this subject.

Flag this post as spam/abuse.

[/collapse]

Posted by Marek Bujnarowski on 01-Oct-2014 03:16

In addition to putting some pressure on Progress Account Managers, I would suggest posting such a requirement to Ideas section (I couldn't find one). It's regularly reviewed by people with sufficient influence. It will also give the whole Community a chance to vote for that, which is the best way to show the impact of missing support for Linux CM, and add appropriate weight to such a request.

Regards

Marek

Posted by Jarek Kowalewski on 01-Oct-2014 12:09

Thanks Vinko & Marek.

Here we go: http://goo.gl/x7U2cA

Vote!

:)

This thread is closed