Our c++ application(32 bit) was deployed in Oracle Enterprise Linux 6.6 and Tuxedo was 11.1.1 (Model was SHM)
We migrated our application to Oracle enterprise Linux 7.7 and Tuxedo to 12CR2(12.1.3 Patch 120)
Now we have enabled Model = MP. SITE1 master and SITE2 slave
We have created two linux virtual machines. Existing application executables are duplicated to both the virtual machines when we enabled MP mode(VM1 – SITE1 & VM2 – SITE2). These two virtual machines may get loaded due to the processes which are outside tuxedo domain(For example, we have C++ jobs, which are outside tuxedo services).
When one of the VM is completely loaded (RAM / CPU utilization is 100%) due to the load from processes which are from outside Tuxedo scope, Tuxedo doesn’t identify this load on the specific VM. It still routes tuxedo service calls to the loaded VM when second VM is completely free. This is causing Tuxedo transactions to be stuck till the RAM is released by the non – Tuxedo process.
In short, in MP mode, when one VM(SITE1) is loaded with non-Tuxedo load, Tuxedo doesn’t do a proper load-balancing. It still routes Tuxedo workqueue requests to the loaded VM(VM1 - SITE1) even if slave machine(SITE2) has negligible load.
This is causing Tuxedo transactions to get stuck and eventually timeout
Could you please review and let me know if this is an issue / Some config changes to be made on tuxconfig?
Sample Tuxedo file and machine configurations are attached.
Thanks
TuxedoIssue.docx (16.89 KB)