My latest aproach to see if MK services are alive was sending MT_PING messages. I'm not 100% sure but, it looks like if MK is down those MT_PING messages (from client that hass been running for a while) got accumulated in zmq messaging layer (king of hwnd buffer?), and than when MK came up, those MT_PING messages flooded MK instance. Looks like this is really not the way how it is planned to be used, but on other hand if it could be made to survive such "dos atack" would be nice...
As long as I do not hve firm and reproducible scenarion to repeat this I do not plan to open issue on guthub. regards Dne ponedeljek, 04. julij 2016 09.16.16 UTC+2 je oseba Michael Haberler napisala: > > > > Am 03.07.2016 um 23:58 schrieb Alexander Rössler > <[email protected]>: > > > > > > Michael Haberler writes: > > > >>> Am 03.07.2016 um 15:02 schrieb Klemen Zhivko <[email protected] > <javascript:>>: > >>> > >>> Experimeting (testsing) with haltalk, mkwrapper on BBB, I now two > times received: > >>> Assertion failed: state == term_ack_sent || state == term_req_sent2 > (pipe.cpp:316) > >> > >> that pipe.cpp is likely part of libzmq > >> > >> taking a guess as to what you are actually doing: you are sending > >> something over a zeromq socket which violates the zmq protocol > > Zhivko, can you take a look at the core dump if haltalk does create > > one. Even if you sent something that does cause problems you should not > > be able to kill the Haltalk process. > > > > PS: this thread should be moved to GitHub > > After it is a proper bug report, this mail is no such thing. > > http://www.machinekit.io/docs/getting-help/#how-to-report-an-error is > pretty clear and I expect people to follow it for basic politeness: you are > expected to explain what you did before you ask for other people's time and > attention. Throwing up an assert line without doing these basics is a waste > of everybody's time. > > > re core dump: things are a bit different - the zeroMQ library takes the > stance that if you send it so much nonsense that it cannot continue it will > fire an assert. While this is a bit harsh, it is pretty effective in > sorting out non-compliant interaction. Therefore, staring at a core dump > will help about zero. > > > > What will help if zhivko prepares a test case which can be easily > reproduced. Everything else is guesswork. So far we do not even know what > he sent, to which socket and how - from which client. > > In particular it would help to create a minimal zeroMQ program, and the > client interaction, which narrows the issue down. > > > Can I therefore ask Klemen to do that homework so we can put aside the > guessing and staring into the crystal ball. Not sure if suggested this 5, > 10 or 20 times to you, for some reason it just does not seem get through: > > 1. follow these steps: -->> > http://www.machinekit.io/docs/getting-help/#how-to-report-an-error > > 2. file an issue _after you have done so_. > > > > >> > >> > >>> > >>> I tried to find pipe.cpp and I cannot find it among sources.. > >>> Maybe somebody know what could be the cause of error and how to > troubleshoot? > >>> > >>> When such error apears MK shutdowns. It doesn-t happen consistently, > so it is hard to reproduce. > >>> > >>> > >>> -- > >>> website: http://www.machinekit.io blog: http://blog.machinekit.io > github: https://github.com/machinekit > >>> --- > >>> You received this message because you are subscribed to the Google > Groups "Machinekit" group. > >>> To unsubscribe from this group and stop receiving emails from it, send > an email to [email protected] <javascript:>. > >>> Visit this group at https://groups.google.com/group/machinekit. > >>> For more options, visit https://groups.google.com/d/optout. > >>> <linuxcnc.log> > > > > > > -- > > Alexander > > -- website: http://www.machinekit.io blog: http://blog.machinekit.io github: https://github.com/machinekit --- You received this message because you are subscribed to the Google Groups "Machinekit" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. Visit this group at https://groups.google.com/group/machinekit. For more options, visit https://groups.google.com/d/optout.
