On Feb 3, 2009, at 11:33 AM, Lars Marowsky-Bree wrote: > Hi, > > I've been thinking about the fact that there's a bunch of scenarios > CTS > isn't good at simulating. > > Essentially, it will reboot nodes, or cause processes to crash, way > too > often for some problems to show. > > From the point of view of the cluster, that is actually not a typical > work load: clusters are expected to stay up for long, and be largely > stable. > > I think we might need an additional selection of tests (and perhaps > one > or two additional tests) which instead simulate what users would do: > perform a lot of switch-overs, standby, cib operations, LRM cleanups, > reprobes, status queries, and only extremely rarely - if at all - > reboot > a node. > > Of course it's still hard to simulate "long-running" as in "a long > time", because, well, that happens to take a long time. ;-) But we can > compress the rest into a shorter period. Thoughts? Comments?
I've added a --loop option to cts so that we can add (and selectively enable) tests like this. I've not had a chance to add any, but there is definitly a case for them to be written. Would you like to kick it off by adding your standby test? Just make sure the test has: def is_applicable(self): return self.CM.Env["loop-tests"] _______________________________________________ Pacemaker mailing list Pacemaker@oss.clusterlabs.org http://oss.clusterlabs.org/mailman/listinfo/pacemaker