Custom Utilities/Applications
Plenty of applications that are not cluster-aware will continue to run in the event of a
failover. Remember, a failover event doesn??™t necessarily mean that a node is physically
failing. It could mean that someone shut down a service without using the cluster service
to do so. A custom, cluster-unaware application could continue to run on the failed
node and start automatically on another node. This could be potentially catastrophic.
CHAPTER 7 n CLUSTERING 182
I worked with a client who had a custom Java application that sent data to a business
partner via an early form of a Web service. It ran on a SQL Server that had centralized data.
When we upgraded that SQL Server to run in a clustered configuration, we had to manually
set up the Java application on each node (it was cluster-unaware). The application
was initiated by a scheduled SQL Agent job. Everything worked fine until someone (OK,
it was me) shut down the SQL Agent on the active node without using the cluster service.
This cascaded into a complete failover, which started SQL Agent on another node, which
in turn started the Java application on another node.
Pages:
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355