Hi All, In my humble opinion, I think we should create a single location for timeouts and the timeouts should be the same for similar operations, eg. restarting a locator. Further, I think that if the timeout is exceeded, then it should throw a very clear exception stating what has happened. This way, if we realize that the server is taking longer to start, we can adjust one number and it will affect all tests. Then, if in a case the server takes too long to start, it will be obvious as well.
Thanks, Mark > On Jul 11, 2018, at 10:05 AM, Udo Kohlmeyer <u...@apache.org> wrote: > > Hi there Dan, > > Whilst 5min seems to be a viable option, I believe that knowing how long an > operation should take and reacting if it doesn't complete in that time is > better than waiting a standard amount of time. I like the faster feedback > option, rather than the standard timeout across the board. Now, that said, > crazy timeouts like 1-10s are maybe a little low. > > Maybe 1min rather than 5min? With exception to disk recovery tests, which > feasibly could take more than 1min. I cannot think of a single operation in > Geode, that should realistically should take more than 60s. > > --Udo > > > On 7/11/18 09:07, Dan Smith wrote: >> Hi all, >> >> We have a bunch of tests that are using awaitility. It seems like every >> tests is picking some random number of it's timeout, usually in the range >> of 10-60 seconds. >> >> I'd like to change all of our tests to use a standard timeout that is much >> higher, to avoid worrying about whether our timeouts are to low. So I >> propose introducing our own GeodeAwaitility class that sets a timeout of 5 >> minutes and replacing all of our usage with that. >> >> -Dan >> >
