----- Original Message ----- > From: Marco Betti <m.be...@gmail.com> > To: Tomcat Users List <users@tomcat.apache.org> > Cc: > Sent: Friday, October 21, 2011 7:18 AM > Subject: Re: FarmWarDeployer undeploy on slave node when tomcat restart > > Hi Keiichi, > many thanks for your suggestion. > By setting synchronous replication everything works. > Unless processing updated war by one message, I think that async replication > shouldn't be used (personal opinion). > By the way I wander If asnyc replication couldn't lead to some other > unexpected behavior. > > Many thanks, > Marco > > > > > On Fri, Oct 21, 2011 at 12:10 PM, Keiichi Fujino <kfuj...@apache.org> > wrote: > >> I checked FarmWarDeployer's code. >> When WAR is updated, FarmWarDeployer sends UndeployMessage and sends >> FileMessage after that. >> The node which received UndeployMessage undeploys war. >> The node which received FileMessage deploys war. >> >> If FileMessage is received while processing UndeployMessage, the >> following error messages will be outputted. >> "SEVERE: Application /example in used. touch war file example.war > again!" >> >> In the case of asynchronous replication (default), FileMessage may be >> received while processing UndeployMessage. >> >> I think that it may be better to process updated war by one message. >> Although I consider correction of a code, there is one work around. >> >> Workaround is change in synchronous mode. >> >> Please try the following. >> <Cluster > className="org.apache.catalina.ha.tcp.SimpleTcpCluster" >> channelSendOptions="6"> >> >> >> 2011/10/21 Marco Betti <m.be...@gmail.com>: >> > Tried also on Tomcat 7.0.22 >> > >> > Same behaviour: when i restart node1, FarmWarDeployer undeploy my war >> both >> > from node1 and node2, but re-deploy it only on node1. >> > If I touch my war in my watchDir on node1, tomcat redeploys it > correctly >> on >> > both nodes... >> > >> > Is tomcat clustering used? >> > >> > If so, does anybody use FarmWarDeployer ? >> > >> > Many thanks, >> > Marco >> > >> > >> > >> > On Thu, Oct 20, 2011 at 6:21 PM, Marco Betti <m.be...@gmail.com> > wrote: >> > >> >> Hi all, >> >> I'm new to the list. >> >> I'm trying to configure tomcat 6.0.33 clustering. >> >> Session replication is ok. >> >> I've a problem with FarmWarDeployer. >> >> I found that: >> >> >> >> 1. You have to place the <Cluster> node under the > <Host> element. >> >> >> >> Tomcat 6.0.x complains that the element Cluster/Deployer element > is not >> >> understood at the Engine level. >> >> >> >> 2. One node needs to have watchEnabled="true", all other > nodes >> >> watchEnabled="false" >> >> >> >> 3. deployDir is the same as appBase >> >> >> >> >> >> and after I found this post that confirmed these points to me: >> >> >> >> http://www.mail-archive.com/users@tomcat.apache.org/msg80374.html >> >> >> >> The problem is the following: >> >> let node1 be the master (watchEnabled="true") and node2 > the slave >> >> (watchEnabled="false") >> >> This is my configuration: >> >> <Deployer > className="org.apache.catalina.ha.deploy.FarmWarDeployer" >> >> >> >> >> > tempDir="/appserv/tomcat/tomcat60/apache-tomcat-6.0.33/server/clu001node01/temp/" >> >> >> >> >> > deployDir="/appserv/tomcat/tomcat60/apache-tomcat-6.0.33/server/clu001node01/webapps/" >> >> >> >> >> > watchDir="/appserv/tomcat/tomcat60/apache-tomcat-6.0.33/server/clu001node01/farm/" >> >> watchEnabled="true"/> >> >> >> >> >> >> If I copy a war to watchDir of node1 evertything works fine and > cluster >> >> deploy is achieved. >> >> >> >> When i restart node1, FarmWarDeployer undeploy my war both from > node1 >> and >> >> node2, but re-deploy it only on node1. >> >> >> >> node1 log: >> >> >> >> INFO: Cluster wide remove of web app /example >> >> Oct 20, 2011 5:50:46 PM org.apache.catalina.startup.HostConfig >> >> checkResources >> >> INFO: Undeploying context [/example] >> >> Oct 20, 2011 5:50:46 PM org.apache.catalina.startup.HostConfig > deployWAR >> >> INFO: Deploying web application archive example.war >> >> >> >> >> >> node2 log: >> >> >> >> INFO: Undeploying context [/example] >> >> Oct 20, 2011 5:50:47 PM > org.apache.catalina.ha.deploy.FarmWarDeployer >> >> messageReceived >> >> SEVERE: Application /example in used. touch war file example.war > again! >> >> >> >> >> >> Obviously, if I touch my war in my watchDir on node1, tomcat > redeploys >> it >> >> correctly on both nodes... but this cannot be the solution... >> >> >> >> >> >> Is it a known problem? >> >> >> >> Could anyone help me? >> >> >> >> Many thanks, >> >> Marco >> >>
I've noticed this as well with FarmWarDeployer. However, I've always used synchronous as the channel option so it's not been a problem. As to the placement of the cluster, yes you have to place it at the Host element in order for FarmWarDeployer to work. That note is buried in the documentation somewhere. Another thing on my to-do list is to write up a basic clustering HOW-TO. Unfortunately right now I have no spare cycles. One of the issues I have with FarmWarDeployer is that upon Tomcat restart the application is undeployed and redeployed cluster-wide. This can have several unintended consequences. 1. Makes the application unavailable cluster-wide (ouch) 2. Slows down restart 3. Could trigger permgen issues for troublesome applications It would be nice if FarmWarDeployer checked file creation times between the deployDir and watchDir. Then if the deployDir times are later than the watchDir times, don't undeploy and redeploy the application. I'm not sure how this would work with versioned web applications (just typing off the top of my head), but this might be a nice enhancement to the FarmWarDeployer. I know . . . patches welcome. Unfortunately I have no spare cycles at present. . . . . just my two cents. /mde/ --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
