On 25/11/2014 03:55, Chris Gamache wrote: > So first, I'd be glad to. Just to be clear-- I'm not in need of a lecture, > but will willingly take one if anyone can help me get down to a solution. > My first attempt at asking for some assistance was met with crickets.
Your first question was very specific: <quote> My question is directed to other tomcat admins out there who are handling this scenario gracefully. What are you doing to handle this problem? </quote> I'd take the lack of response as meaning no other admins are handling this scenario gracefully. That doesn't surprise me the feature isn't that widely used. > I > haven't figured out why when I post (what I think are) well written, well > researched and discussion-provoking questions they get no traction It got no traction as you asked a specific question that I suspect no-one had an answer to. > but if I > cry "the sky is falling" folks are more willing to attempt to lend a hand. Because you are asking much broader, less specific question. You have gone from "How are other folks handling this?" to "Is this the expected behaviour? Is there a bug here?". Those are very different questions. > Here's my go at attempt #3: > > I have a tomcat cluster. It is set up with much care according to > http://tomcat.apache.org/tomcat-7.0-doc/cluster-howto.html. > > It mostly works fine. On occasion (twice a week or so) there will be one or > more servers which didn't get the message that a new war was deployed > (continuous deployment using the tomcat parallel deployment scheme. e.g. > theapp##007.war) and they happily continue to run the old version of the > war. > > In a farm deployment scenario, the master node will announce to the cluster > a new artifact is available and then the clustered tomcats will retrieve > and deploy the new artifact. I can't pin down the problem, but let just say > for argument's sake it is a true link-down situation. There doesn't seem to > exist a mechanism to re-announce, or announce at regular intervals. You could always check the source code to be sure. From memory, it should be fairly obvious from the message types whether or not such a feature exists. > This seems like a real weakness in the scheme. Agreed. > That makes me think I'm missing something obvious. Probably not. > If it works like it says it should in the docs I > shouldn't be having this issue. The docs don't appear to cover this scenario. > Either there's something wrong with my > config or there's a problem with tomcat. Based on the hair I have left > after pulling mine out, I'm leaning toward a problem with tomcat. Is anyone > leveraging FarmWebDeployer and Parallel Deployment in their architecture > successfully? If so, DenverCoder9, what did you see?!? I know the parallel deployment works with the FarmWarDeployer since I tested that when I wrote the parallel deployment implementation. I didn't test behaviour with failed nodes. Mark --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
