-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Chip,
On 7/26/12 3:54 PM, Chip McVey wrote: > I have read that Tomcat does not unload servlets (even though it > would be permissible to do so), but as you say, I have done some > logging in the destroy method, and I see this behavior. Can you give us a stack trace of that event? > A certain servlet within app B & C (let's call them B1 and C1) upon > init make a call to a servlet in app A (let's call it A1). What I > see is A1, B1, and C1 all initialize. The init is done by a thread > called " pool-2-thread-1", which seems like a system thread. Ok no > problem. > > Then, for some reason, A1's destroy method is called, and it is > called by a thread named " http-apr-8080-exec-3", making me think > it is being called by a thread processing the request for either B1 > or C1, since no other requests from anyone outside of B or C have > happened yet. http-apr-8080-exec-3 should be a request-processing thread, which has no business calling destroy() on a servlet. > After destroy winds down, then, A1's init method is called by a > thread named " http-apr-8080-exec-11" - another request processing > thread I assume. Yes. This is probably happening because the servlet needs to be brought (back) into service. > Everything all works out in the end, and after that I don't see > A1, B1, or C1 ever destroyed - but am I guaranteed that they won't > ever be destroyed again unless TC is shutdown or they are manually > unloaded? There are no guarantees, but Tomcat shouldn't randomly unloading your servlets. Do you only see one of these (surprising) unloads, or do they happen a lot over time? > I understand that given servlets can by spec go up & down, maybe > using them was not the right choice for an application that needs > permanency, but we're too late in development to make that switch > now. Is your webapp somehow being damaged by these unload operations? A servlet should be fairly lightweight, and the init method "should" run very quickly. > So I'm looking for a guarantee that no automatic destroys will > happen going forward just because Tomcat decides to do it on its > own. You will never get that guarantee, because the servlet spec does not require it. > If it is instructed to shutdown/undeploy/unload by some outside > system or by a human user, that's fine, I'm not worried about > that. But I don't want these servlets to be destroyed just because > Tomcat decides to do so for whatever reason. Sorry, you can't control that. What you can do is architect your application in such a way that it doesn't matter. - -chris -----BEGIN PGP SIGNATURE----- Version: GnuPG/MacGPG2 v2.0.17 (Darwin) Comment: GPGTools - http://gpgtools.org Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/ iEYEARECAAYFAlASqKAACgkQ9CaO5/Lv0PCcAQCfTz5FSmH+dangfTdWWu0urO2T a+wAoMRX9W77BTrJMphN6OJUKTBsVVsL =Si2b -----END PGP SIGNATURE----- --------------------------------------------------------------------- To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org For additional commands, e-mail: users-h...@tomcat.apache.org
