Thanks, Jeff. And just to clarify - this will *try* to do a graceful shutdown first, wait for the timeout period (10 seconds) and then force it if that was unsuccessful. The previous behavior was to wait for 5 seconds, and then give up entirely.
We'll have to take another look at this when we move to a newer tomcat package. I have some ideas on how to improve this further - like switching to a systemd unit rather than init script for el7, and putting tunables in an external config (i.e. /etc/sysconfig/tomcat) like the tomcat RPMs generally do. __Jason On Wed, Jul 12, 2017 at 11:44 AM, Jeff Elsloo <[email protected]> wrote: > I discussed this with Jason, reviewed the PR and will be merging it > soon unless someone has concerns. I asked specifically about "force" > being the default shutdown mode, and that was done intentionally. > There might be a use case for a graceful shutdown with typical > applications deployed into Tomcat, but Traffic Router does not service > any long running sessions, so getting it shut down quickly is actually > desired. > > We can use this new init script and make changes as necessary in the > future, but this should be an improvement. Hopefully we won't have to > `kill -9 <pid>` anymore. > -- > Thanks, > Jeff > > > On Tue, Jul 11, 2017 at 3:37 PM, Jason Tucker <[email protected]> > wrote: > > FYI - opened ticket and PR for this issue: > > > > The tomcat init script has a few problems: > > > > 1. "Clean" shutdowns frequently timeout, and the scripts give up, leaving > > tomcat running > > > > 2. Normal tomcat shutdown actually involves spinning up a second jvm > > instance. Right now, we start this second instance with the same > > CATALINA_OPTS as traffic_router, which can be problematic on > > memory-constrained hosts. > > > > https://issues.apache.org/jira/browse/TC-416 > > https://github.com/apache/incubator-trafficcontrol/pull/724 > > > > Thanks, > > > > _Jason >
