Re: NDB engine possibly not a realtime database engine?

[Stefan Auweiler]

Thanks for the reply, Mikael
we'll try to log and tweak NTP, to catch this.


How could I prove that this is coming from the OS?
Is there a call 'give me CPU' which is not answered for a while?
Is the log message a 100% prove of this?

Customer is really complaining about real time capabilities here.


We never noticed this message
2017-03-16 11:10:59 [ndbd] WARNING -- Ndb kernel thread 10 is stuck in: Job Handling elapsed=6336
on two systems in parallel.

If the one node doesn't respond within ms (as usual) why is the other nod of the node group not delivering instead?



BTW: we did not see this in Solaris 10 with NDB 7.3.7 but had to change binaries do go to Solaris 11, so we went to NDB 7.4.9.
Have there been any changes in this area?



Thanks
Stefan