| Dave |
Mon Nov 12 18:17:18 GMT 2007 |
| |
Two questions/suggestions for Primrose 3
1. Have you considered using SLF4j (http://www.slf4j.org/)as your logging interface?
2. Is it necessary for Primrose to open a connection in each pool at startup? This can significantly delay (or hang) servers that are hosting lots of applications with many pools. Can this be a configuration option? |
| |
| guest |
Tue Nov 13 18:36:52 GMT 2007 |
| |
1. We try to stay away from 3rd party logging tools, and any 3rd party APIs in general, because it builds in what we regard as non-essential dependancies. We may in the future consider providing the ability to use one though, as this is the 2nd request for such a feature.
2. I don't really feel that opening a single connection at startup should really "hang" and application on startup, but we could add a parameter I suppose that indicates if you wish for connection to be opened. |
| Dave |
Tue Nov 13 20:46:58 GMT 2007 |
| |
1. I can appreciate keeping the dependencies to a minimum. It was nice being able to recompile primrose without hunting down 50 other jar files. But if you decide that the advantages of SLF4j offset the bother of compiling against a standard api, I recommend it specifically because it is quite powerful and able to integrate with any logging implementation (log4j, commons-logging, etc. or even your own).
2. I experienced the hanging at startup when I installed primrose in a dev environment with some misconfigured pools (non-existent addresses and such). Probably socket timeouts would occur eventually to allow startup to proceed. And it was easy enough to find the problem pools in the logs but in our production environment where we have quite a few applications per vm we don't want a slow/broken database or misconfigured pool config to hold-up startup for all apps.
And I'd like to add that I think Primrose has a remarkably clean implementation. (maybe due in part to a lack of external |
|
