hi Yuri, thanks a lot for posting the workaround to the mailing list! I am sure it will be very useful for people who will run into a similar problem. kind regards, risto
2016-08-04 17:59 GMT+03:00 Yuri Kazankin <yuni...@yahoo.com>: > Hello! > > Risto, Thanks for the you tips, the problem was solved. For systemd in > service settings added “KillMode=process”, this makes it possible to kill > only the basic process and does not affect the child processes. > > $ cat /lib/systemd/system/sec@.service > [Unit] > Description=Simple Event Correlator (instance %I) > After=syslog.target > > [Service] > Type=forking > PIDFile=/run/sec-%I.pid > ExecStart=/usr/bin/sec -detach -pid=/run/sec-%I.pid -conf=/etc/sec/%I.sec > --dump=/tmp/sec-tomcat.dump --nochildterm -log=/var/log/sec-%I --intevents > $OPTIONS_%I > EnvironmentFile=/etc/sysconfig/sec > *KillMode=process* > > [Install] > WantedBy=multi-user.target > > > > -- > WBR > Yuri Kazankin > yuni...@yahoo.com > > On 2 авг. 2016 г., at 0:53, Risto Vaarandi <risto.vaara...@gmail.com> > wrote: > > hi Yuri, > if your platform is Linux, you can figure out the parent process ID by > running "ps -ef" and inspecting the third column. This command lists all > processes on your system, so you would probably need to grep the output for > relevant processes (e.g., ps -ef | grep tomcat). When you shut down sec, > please also inspect its log, because sec reports the ID of every process it > sends the TERM signal. The relevant sec log messages look like follows: > SIGTERM received: shutting down SEC > Sending SIGTERM to process 20879 > > Nevertheless, after having a look into the scripts, I have some doubts if > the application is running as sec child process, since from the shell > process that runs restart-tomcat.sh you are invoking startup.sh which > creates another process (that process should become a grandchild of sec > which no longer gets SIGTERM). Since you are experiencing this problem when > using systemctl command for restarting sec, maybe this issue is related to > how systemd shuts down a service. After browsing a web, I run into couple > of threads that describe a similar scenario, for example: > http://stackoverflow.com/questions/32208782/cant- > detach-child-process-when-main-process-is-started-from-systemd > > In order to find out if you are experiencing a similar problem, you could > try stopping sec in two ways -- first by shutting it down manually by > sending TERM signal to sec process, and then stopping it with 'systemctl > stop sec@tomcat'. If your application stays alive in the first case but > dies in the second case, it has to be systemd which stops your application > alongside with sec, probably because your application was started by sec > and belongs to the same process group. Nevertheless, if systemd uses the > TERM signal for taking down your application, you could still use the trap > '' 15 statement in the restart-tomcat.sh script for ignoring this signal. > However, I would recommend to create systemd service files for your > application, so that you could start and stop it with systemctl. Invoking > "systemctl restart ..." instead of restart-tomcat.sh would probably solve > the problem, since systemd is made aware that your application is a > separate service. > > kind regards, > risto > > 2016-08-01 14:14 GMT+03:00 Yuri Kazankin <yuni...@yahoo.com>: > >> Hi Risto, >> >> Sec process is seen as: >> /usr/bin/perl -w /usr/bin/sec -detach -pid=/run/sec-tomcat.pid >> -conf=/etc/sec/tomcat.sec --debug=6 --dump=/tmp/sec-tomcat.dump >> --nochildterm -log=/var/log/sec-tomcat --nointevents -input=/var/log/tomcat/ >> catalina.out >> >> Unfortunately, --nochildterm option does not fix the problem. I cannot >> find the dump file, I tried to specify it in the start script >> (--dump=/tmp/sec-tomcat.dump), but without result. >> >> The command "ps wuaxf" does not reveal how the child process: >> # ps wuaxf >> root 31114 0.1 0.2 150232 9724 ? S 13:42 0:01 >> /usr/bin/perl -w /usr/bin/sec -detach -pid=/run/sec-tomcat.pid >> -conf=/etc/sec/tomcat.sec --debug=6 --dump=/tmp/sec-tomcat.dump >> --nochildterm - >> root 31214 4.8 6.8 3756928 266000 ? Sl 13:44 0:39 >> /opt/java-8-sun/jre/bin/java -Djava.util.logging.config. >> file=/opt/tomcat-7.0.55/conf/logging.properties >> -Djava.util.logging.manager=org.apache >> >> In my restart-tomcat.sh: >> kill -9 `cat /run/tomcat-7.0.55.pid` ; sleep 5 ; /opt/tomcat-7.0.55/bin/ >> startup.sh >> >> In the attachment native scripts of Apache Tomcat: startup.sh and >> catalina.sh >> >> Trap will try if nothing else helps. >> >> Can you something else advise? >> >> >> >> >> >> -- >> WBR >> Yuri Kazankin >> yuni...@yahoo.com >> >> On 31 июля 2016 г., at 22:23, Risto Vaarandi <risto.vaara...@gmail.com> >> wrote: >> >> hi Yuri, >> my apologies for not answering earlier -- your e-mail landed in my spam >> folder initially :( >> >> I am suspecting that your application is started in a way which makes it >> a child process of sec. However, when sec shuts down, it sends the TERM >> signal to all its child processes by default. If the shell process that is >> executing the content of /opt/scripts/restart-tomcat.sh is still running, >> this process will get the TERM signal as well. Therefore, if the >> application was started with the shell 'exec' directive that replaces the >> shell process with the application process, it is actually your application >> which gets the TERM signal. >> >> In order to find out whether your application is running as a child >> process of sec, send the USR1 signal to the sec process which creates the >> dump file (by default located at /tmp/sec.dump). This file contains >> detailed information about internal state of sec, and also provides a list >> of all child processes. >> >> If your application is indeed stopping because of the TERM signal from >> sec, there are couple of workarounds: >> 1) start sec with the --nochildterm command line option which disables >> sending TERM to child processes, >> 2) include a 'trap' statement in /opt/scripts/restart-tomcat.sh that >> would ignore the TERM signal (e.g., trap '' 15) >> >> Also, if your application is terminating for other reasons than TERM >> signal from sec, can you share the content of /opt/scripts/restart-tomcat. >> sh? >> >> Hope this helps, >> risto >> >> >> >> 2016-07-29 13:55 GMT+03:00 Yuri Kazankin <yuni...@yahoo.com>: >> >>> Hello! >>> >>> I have a problem with sec 2.7.10 >>> >>> I created the instance: >>> # cat /etc/sysconfig/sec >>> OPTIONS_tomcat="-input=/var/log/tomcat/catalina.out” >>> >>> >>> and configured via systemd >>> # cat /lib/systemd/system/sec@.service >>> [Unit] >>> Description=Simple Event Correlator (instance %I) >>> After=syslog.target >>> >>> [Service] >>> Type=forking >>> PIDFile=/run/sec-%I.pid >>> ExecStart=/usr/bin/sec -detach -pid=/run/sec-%I.pid >>> -conf=/etc/sec/%I.sec --debug=6 --nochildterm -log=/var/log/sec-%I >>> --nointevents $OPTIONS_%I >>> EnvironmentFile=/etc/sysconfig/sec >>> >>> [Install] >>> WantedBy=multi-user.target >>> >>> >>> My rules restart the application after the event in the log: >>> # cat /etc/sec/tomcat.sec >>> type=SingleWithSuppress >>> ptype=RegExp >>> pattern=(^.+OutOfMemoryError.+$) >>> desc=Tomcat restarted after OutOfMemoryError Exception >>> action=shellcmd /opt/scripts/restart-tomcat.sh >>> window=300 >>> >>> And it works great! But after the event, if I try to execute a restart >>> of the instance (systemctl restart sec@tomcat), my application stops((( >>> I can not understand why this is happening and how to fix it ... >>> >>> >>> -- >>> WBR >>> Yuri Kazankin >>> yuni...@yahoo.com >>> >>> >>> ------------------------------------------------------------ >>> ------------------ >>> _______________________________________________ >>> Simple-evcorr-users mailing list >>> Simple-evcorr-users@lists.sourceforge.net >>> https://lists.sourceforge.net/lists/listinfo/simple-evcorr-users >>> >> >> >> >> > >
------------------------------------------------------------------------------
_______________________________________________ Simple-evcorr-users mailing list Simple-evcorr-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/simple-evcorr-users
