While testing High Availability I want to simulate a failing primary server. Killing the process via task manager would work, but I just wanted a more elegant solution using plain SQL. |
To terminate a server using a SQL statement, you might use xp_cmdshell() together with a command that kills a process (e.g. taskkill on Windows). The issue that caused the crash with the above CREATE VIEW statement will be fixed in build 11.0.1.2367. |
OS: Ubuntu10.4 AMD64 / Debian5 AMD64; SQL Anywhere: SA11.0.1.2645 64bit for shut down server, run query in java isql: select * from test_table where id=1; select * from test_table where id=2; select * from test_table where id=1 select * from test_table where id=2 ps: case id 11695638 Is the semicolon deliberately omitted in the 3rd row? That might turn this into a Transact-SQL statement batch (i.e. it will return 2 result sets) - though a crash then would surely look like a bug...
(05 Nov '11, 19:02)
Volker Barth
|
IMO the BEST tests are (a) pulling the power cord and (b) pulling the network caple. The most FUN tests involve water hoses, firearms and so forth... but I never get to have any fun.
...network cable...
... I was thinking about a wood-chopper's axe, but that wasn't in the budget (neither a new server)
some of us were around back when Novell dramatically demonstrated their high availability server at a trade show... with 20 workstations scrolling data from the server, the salesman used an axe to chop the rope suspending a HUGE blacksmith's anvil over one server. There was crash, bang, smoke and fire. And the workstations breifly paused while failing over, then resumed full speed scrolling. Pretty hard to top that demo!