Speaking of monitoring tools and if we don't want to go for third party tools, we can also use VisualVM that comes bundled with Oracle JDK. It can connect to the remote VM (OFBiz process) and start displaying various information.
Very minimal configuration is needed in the form of VM argument to allow for remote monitoring. Also, to enable further analysis of what went wrong, why JVM crashed etc, we should also dump heap as the JVM shuts down. Too many ways and too many options. Probably need to reach a unanimous decision, IMO. Thanks and Best regards, Girish Vasmatkar On Fri, Aug 24, 2018 at 4:56 PM Jacques Le Roux < jacques.le.r...@les7arts.com> wrote: > Thanks Michael, > > Best idea so far! > > Jacques > > > Le 24/08/2018 à 11:08, Michael Brohl a écrit : > > We are monitoring our OFBiz instances with JMX and self hosted Zabbix > [1]. > > > > Zabbix gives you a nice overview about the system health and metrics > like memory consumption etc. It also sends out warnings (Email, SMS or > else) > > if metrics are exceeded (like CPU load or memory consumption) as well as > the system is not accessible. > > > > Looks like this: [2] > > > > There is no programming needed, just some configuration for JMX and > Zabbix. > > > > [1] https://www.zabbix.com/ > > [2] > https://www.ecomify.de/wp-content/uploads/2018/08/Zabbix_Monitoring.png > > > > If we want to see why the demos crash, it might be useful. If we only > want to monitor if the system is up, a simple cron job which sends a mail > > might be enough... > > > > Regards, > > > > Michael Brohl > > ecomify GmbH > > www.ecomify.de > > > > > > Am 24.08.18 um 10:07 schrieb Taher Alkhateeb: > >> Okay all neat ideas, I'm not sure if the energy you will put into > something > >> like this is equal to the value produced but if you want to make this > >> happen I would be happy to assist. > >> > >> How much time will it take to make something like this happen? I ask > >> because it seems Jacques ia getting annoyed with these crashes and we'd > >> like to help him out. > >> > >> On Fri, Aug 24, 2018, 10:59 AM Girish Vasmatkar < > >> girish.vasmat...@hotwaxsystems.com> wrote: > >> > >>> Hi Taher > >>> > >>> Please see my reply below in-line. > >>> > >>> On Fri, Aug 24, 2018 at 12:22 PM Taher Alkhateeb < > >>> slidingfilame...@gmail.com> > >>> wrote: > >>> > >>>> Hi Girish, inline... > >>>> > >>>> On Thu, Aug 23, 2018, 7:25 PM Girish Vasmatkar < > >>>> girish.vasmat...@hotwaxsystems.com> wrote: > >>>> > >>>>> I had earlier replied to this thread but looks like the email did not > >>> go > >>>>> through. I had leaned towards using the tool (only just) instead of > may > >>>> be > >>>>> having a CRON job or an alternative. > >>>>> > >>>>> What I feel now is that may be we can use JMX here and try to use > >>> various > >>>>> in build MBeans that provide CPU usage for the system and also for > the > >>>> JVM > >>>>> process we are concerned about that is OFBiz instance. We should also > >>> be > >>>>> able to get the memory usage of the JVM and if reaches a particular > >>>>> threshold we can be notified. > >>>>> > >>>> Do you have a PoC for all of this? > >>>> > >>> GV : I can have one ready; and there is going to be much doing > involved. > >>> > >>>>> In addition, I think we already add a shutdown hook to the JVM > >>>> process... I > >>>>> am not sure and have not used it much but may be we can use it to > send > >>>> some > >>>>> notifications? Of course, it is applicable for graceful exits of JVM > >>> only > >>>>> and if you just happen to kill the process it won't be of much help. > >>>>> > >>>> The shutdown hook is used for shutting down. I'm not sure what is the > >>>> purpose of mentioning it here? > >>>> > >>> GV : The reason I mentioned shutdown hook was it can be used to > send > >>> notification (may be email) or anything per our needs indicating that > the > >>> demo process was shut down. Per my understanding, shutdown hook > gets > >>> called whenever JVM shuts down gracefully. Graceful word is very > important > >>> here because we won't be able to do much if someone just kills the > process. > >>> The only thing a shutdown hook will add to this is that we will be > notified > >>> then and there. > >>> > >>>>> Hope it makes sense and correct me if I am wrong. > >>>> Well I'm struggling a bit. I didn't understand exactly what needs to > be > >>>> done? I see mixed topics about JMX, Mbeans, Memory monitors and > shutdown > >>>> hooks. First this seems to be more like coding than a tool, and > second I > >>>> have no idea how you want to implement this? > >>>> > >>> GV: Yes, it would mostly be coding rather than being a substitute > for > >>> the tool. My idea was that to have a timer service run within the JVM > and > >>> it access various MBeans for the CPU usage and Memory usages just for > our > >>> monitoring purpose and raise an alert if it reaches a threshold. It was > >>> just to have a glance over how JVM is performing. The disadvantage? The > >>> service will run in OFBiz JVM and there will be considerable amount of > >>> coding involved. > >>> > >>>> My idea for example is simple: create a cronjob that checks the system > >>>> periodically and if the demo process stopped, restart it (or maybe > >>> rebuild > >>>> and restart). To go with your suggestion we need to perhaps first > >>>> understand it. > >>>> > >>> GV: There is nothing wrong with creating a CRON job, per se. The > only > >>> reason why I introduced MBeans in the mix was to be able to sort of > having > >>> OFBiz monitor itself within it's realm, hence use of MBeans. I believe > a > >>> CRON will be able to do it as well. I probably did not get that we > probably > >>> want something that take some action after the JVM has crashed and not > >>> having something that monitors the process and alerts concerned parties > >>> that the process is occupying more than say 2 GB or it's CPU usage has > >>> spiked above 80%. > >>> > >>> All in all, I feel we should choose the solution based on what we want > to > >>> do and whether we want to take it further as well. I do not know what > the > >>> tool does now or whether it can build the system again and restart it > >>> automatically. I also do not know what measures we take in such an > event. I > >>> agree CRON will be simplest of them all, but if the tool provides all > of > >>> these (be able to take corrective measures) and not just send > >>> notifications, then it can also be worth it's salt. Yes, CRON will be > more > >>> technical way of achieving :) > >>> > >>> Thanks and Best regards, > >>> Girish Vasmatkar > >>> HotWax Systems > >>> > >>>>> Best regards, > >>>>> Girish Vasmatkar > >>>>> HotWax Systems > >>>>> > >>>>> > >>>>> On Thu, Aug 23, 2018 at 8:48 PM Jacques Le Roux < > >>>>> jacques.le.r...@les7arts.com> wrote: > >>>>> > >>>>>> Le 23/08/2018 à 14:04, Taher Alkhateeb a écrit : > >>>>>>> I'm not sure why you're hanging this on me, > >>>>>> Because you answered to the bait ;) > >>>>>> > >>>>>>> but sure I'm willing to > >>>>>>> help. > >>>>>> Thanks, much appreciated! > >>>>>> > >>>>>>> Can I get some information on how the crashes are happening and > >>>>>>> how you're getting notified, and I will take it from there. > >>>>>> I think after a crash it's mostly to use dumps there (we have > several > >>>>> from > >>>>>> the recent pas) but I'm not sure they will help, and it takes time > to > >>>>>> analyse. > >>>>>> > >>>>>> In the past I took the time to analyse some of them and it was > >>>>>> interesting. For instance in 2010 I found a bug in a Java version we > >>>> were > >>>>>> using and it > >>>>>> helped me in a custom project I was also doing then: > >>>>>> https://markmail.org/message/byu2ivjn7wckayzz > >>>>>> > >>>>>> Lastly it was mostly lack of memory, despite having 8GB now. I > >>> created > >>>>>> https://issues.apache.org/jira/browse/INFRA-16780 for that, but not > >>>> sure > >>>>>> it was > >>>>>> the reason. At least we have less issues since. > >>>>>> > >>>>>> Before (months ago) the Infra was monitoring our demos and alerting > >>> us > >>>> by > >>>>>> mail (you just had to subscribe). Unfortunately we are on our own > for > >>>>> that > >>>>>> now, too much projects in the ASF... > >>>>>> As as I said initially in this thread I'm currently using > >>>> montastic.com > >>>>>> for the email alerts. > >>>>>> My idea when I started this thread was that it all depends on me, > and > >>>>>> that's bad. So I wanted people to be aware, you are much welcome. > >>>>>> > >>>>>> Jacques > >>>>>>> On Thu, Aug 23, 2018 at 2:29 PM Jacques Le Roux > >>>>>>> <jacques.le.r...@les7arts.com> wrote: > >>>>>>>> Yes we can, will you? > >>>>>>>> > >>>>>>>> Jacques > >>>>>>>> > >>>>>>>> > >>>>>>>> Le 22/08/2018 à 19:29, Taher Alkhateeb a écrit : > >>>>>>>>> Well, we can ask Infra for help, we can check available > >>> solutions, > >>>> we > >>>>>>>>> can create a CRON script that checks things periodically, there > >>> are > >>>>>>>>> multiple ways to go about this. > >>>>>>>>> > >>>>>>>>> My personal preference is for a simple CRON script that takes > >>> care > >>>> of > >>>>>> this. > >>>>>>>>> On Wed, Aug 22, 2018 at 8:25 PM Jacques Le Roux > >>>>>>>>> <jacques.le.r...@les7arts.com> wrote: > >>>>>>>>>> So you prefer that I'm the only one to take care of the demos > >>> and > >>>>> act > >>>>>> on alerts? > >>>>>>>>>> Jacques > >>>>>>>>>> > >>>>>>>>>> > >>>>>>>>>> Le 22/08/2018 à 18:53, Taher Alkhateeb a écrit : > >>>>>>>>>>> I prefer not to include any tools without proper analysis and > >>>>>>>>>>> discussion first. Less is more. > >>>>>>>>>>> On Wed, Aug 22, 2018 at 5:31 PM Jacques Le Roux > >>>>>>>>>>> <jacques.le.r...@les7arts.com> wrote: > >>>>>>>>>>>> Hi, > >>>>>>>>>>>> > >>>>>>>>>>>> Should I consider no answers as a lazy consensus and should I > >>>> send > >>>>>> (rare) alerts to this ML? > >>>>>>>>>>>> Without any answers I'll consider it a lazy consensus in 2 > >>> days. > >>>>>>>>>>>> Jacques > >>>>>>>>>>>> > >>>>>>>>>>>> > >>>>>>>>>>>> Le 17/08/2018 à 12:22, Jacques Le Roux a écrit : > >>>>>>>>>>>>> Le 13/08/2018 à 18:21, Jacques Le Roux a écrit : > >>>>>>>>>>>>>> Le 12/08/2018 à 11:26, Jacques Le Roux a écrit : > >>>>>>>>>>>>>>> Hi, > >>>>>>>>>>>>>>> > >>>>>>>>>>>>>>> This morning I noticed the old demo was down and restarted > >>> it > >>>>>> after cleaning things. > >>>>>>>>>>>>>>> Previously (still some weeks ago) Daniel Gruno's (from > >>> Infra > >>>>>> team) company was kindly providing us a mean to monitor our demos > but > >>>> it > >>>>>> seems that > >>>>>>>>>>>>>>> this mean is no longer available > >>>>>>>>>>>>>>> > >>>>>>>>>>>>>>> I have asked about it and will let you know about it... > >>>>>>>>>>>>>>> > >>>>>>>>>>>>>>> Have a good weekend > >>>>>>>>>>>>>>> > >>>>>>>>>>>>>>> Jadques > >>>>>>>>>>>>>>> > >>>>>>>>>>>>>> Daniel confirmed it's terminated. I turned to UpTimeRobot > >>>> which > >>>>>> is free and seems as well good :) > >>>>>>>>>>>>>> Jacques > >>>>>>>>>>>>>> > >>>>>>>>>>>>>> > >>>>>>>>>>>>> This thread started on user ML but I don't want to bother > >>>>> everyone > >>>>>> with technical details. > >>>>>>>>>>>>> I used my own @a.o email to create the monitoring. > >>> UpTimeRobot > >>>> is > >>>>>> certainly the best free monitoring tool, with some possibilities > >>> others > >>>>>> don't give. > >>>>>>>>>>>>> But the free version has an inconvenient. You can only check > >>>>> every > >>>>>> 5 mins and when the instances restart it takes more than 5 mins > each. > >>>>>>>>>>>>> So everyday I get a down an up alerts for each. I have > >>> switched > >>>>> to > >>>>>> montastic.com. > >>>>>>>>>>>>> I was wondering if we don't want to share that here. > >>>>>>>>>>>>> We could then have these alerts here and any committer, using > >>>> the > >>>>>> info inhttps://svn.apache.org/repos/asf/ofbiz/tools/demo-backup > >>> could > >>>>>> handle issues. > >>>>>>>>>>>>> It seems better, isn'it? > >>>>>>>>>>>>> > >>>>>>>>>>>>> Jacques > >>>>>>>>>>>>> > >>>>>>>>>>>>> > > > > > >
