Hi again Rainer, It seems some of our software updates must have solved the issue, as it's been working perfectly now for several weeks. I think we may conclude that it didn't had to do with the the clustering implementation. I feel sorry that you put time into patching the issue. In case the problem ever comes up again, the first thing I'll do is test the patch you've sent.
Thanks a lot. 2009/10/14 Bocalinda <bocali...@gmail.com> > Hi Rainer, > > I'm sorry for the long delay in getting back to you. > I got sidetracked by another issue and wasn't able to look into this > earlier. > > However, the bad news is that the administrators didn't allow me to apply > the patch as it is a production server. > I can apply it to pre-production, but the error isn't reproducable there... > > 2nd however; It's been 12 days since we've seen the listenerStart error due > to the classDefNotFound (in production). > I need to check with the developers whether they accidentily made any > changes that might have solved this problem. > It might as well be that we have been lucky and that the error starts > appearing again tomorrow. > > I will try to reproduce the problem in another server, in order to try out > the patch you supplied. > > Thanks again. > > > 2009/10/3 Bocalinda <bocali...@gmail.com> > >> Wow, that's what is called a quick turn around! :) >> >> Thanks. I'll try as soon as possible and let you know. >> >> 2009/10/3 Rainer Jung <rainer.j...@kippdata.de> >> >> On 03.10.2009 18:11, Bocalinda wrote: >>> > Hi Rainer. >>> > >>> > <wild guess> >>> >> I remember problems with clustering which were of the following kind: >>> >> session replication data was received on a node although the context >>> >> wasn't yet fully initialized. >>> >> >>> >> Maybe we run into classloader troubles in that case? >>> >> >>> >> This could only happen if another node in the cluster is running all >>> the >>> >> time between stop and restart of the node, which logs the exception. >>> >> >>> >> This could certainly be the case as I have 4 nodos in cluster. I >>> > stop/restart one by one, in order to ensure the availability of the >>> webapp. >>> > >>> > >>> >> How easy can your reproduce the problem? In case I provide a patch, >>> >> could you test, whether the problem goes away? >>> >> </wild guess> >>> >> >>> >> >>> > Taking into account that this is a production server, it won't be easy >>> to >>> > play with it, but I can propose to take 2 nodes out of the cluster to >>> do >>> > testing with. >>> >>> You can find a patched catalina-ha.jar and in case you are interested >>> also the respective source code file at >>> >>> >>> http://people.apache.org/~rjung/patches/tc6-cluster-context-startup-order.zip<http://people.apache.org/%7Erjung/patches/tc6-cluster-context-startup-order.zip> >>> >>> I did a one to one port from TC 5.5 to TC 6.0. So please test carefully >>> and let us know the results, so we can include the patch into the >>> standard code in case it proves it helps. >>> >>> Regards, >>> >>> Rainer >>> >>> --------------------------------------------------------------------- >>> To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org >>> For additional commands, e-mail: users-h...@tomcat.apache.org >>> >>> >> >
