Multiple-Nodes javabatch deployment about fhir HOT 5 CLOSED

While working on #633 I thought we should probably add an array to "batch-uri": "https://localhost:9443/ibm/api/batch/jobinstances" so it could work even if a specific node it was pointed to failed.

from fhir.

While working on #633 I thought we should probably add an array to "batch-uri": "https://localhost:9443/ibm/api/batch/jobinstances" so it could work even if a specific node it was pointed to failed.

I have just created issue #636 for simplify the batchclient and configuration.

from fhir.

I have run multiple deployment tests in my local and in CDT, I can not reproduce the job submission delay issue which happened during Connectathon 22.
Configuring separated job schedulers and executors makes the configuration and workflow more complicated - have to add new configurations and the job can only be submitted to the fhir server instance with scheduler configured.

So, I would prefer to continue our current simple deployment architecture to allow each fhir server instance to act as executor for its received job request, and leverage load balance to send the request to the alive server.

Test 1:
In local env, run multiple instances of fhir servers which share the same local DB2 javabatch database and without separation of job scheduler and Job executor. test jobs were submitted in few seconds.

Test 2:
In local env, run multiple instance of fhir servers which share the same IBM Cloud DB2 javabatch database and without separation of job scheduler and Job executor. Test jobs were submitted in few seconds.

Test 3:
In CDT env, run multiple instance of fhir servers which share the same IBM Cloud DB2 javabatch database and without separation of job scheduler and Job executor. Test jobs were submitted in few seconds.

Test4:
In local env, run multiple instance of fhir servers which share the same local DB2 javabatch database and with MQ and separated job schedulers and Job executors. Test jobs were also submitted in few seconds.

Test5:
In local env, run multiple instance of fhir servers which share the same IBM Cloud DB2 javabatch database and with MQ and separated job schedulers and Job executors. Test jobs were also submitted in few seconds.

from fhir.

While working on #633 I thought we should probably add an array to "batch-uri": "https://localhost:9443/ibm/api/batch/jobinstances" so it could work even if a specific node it was pointed to failed.

I have just created issue #636 for simplify the batchclient and configuration.

I investigated this approach in my PR. It does not work as you'd think. The start method actually starts the job on the local Jar.

from fhir.

Closed in discussion

from fhir.