We’re running 6.10 CE, but want to implement a new feature to ONE. So we wanted to test the pipeline of working, and tried building ONE without any changes to it first. So we took down the gitlab sources, checked out 6.10-ce, built the sources and installed targets in a custom directory, and only picked the executable ‘oned’ (as it will contain our changes). If we replace the original ‘oned’ with this new one, ‘systemctl start opennebula’ will timeout. We can’t see anything in the logs that actually goes wrong though. And if we launch manually all other services first (opennebula-novnc, opennebula-scheduler, opennebula-hem, opennebula-sunstone), and then manually launch ‘oned -f’, it actually stays up and we can do “some” testing on the GUI. But sadly even though virtual machines get created on the GUI, they just stay at PROLOG and never really launch. What’s going on here? Why can’t we replace the 6.10 ‘oned’ executable with one we build ourselves?
If it’s to any help, gdb says it’s stuck here:
#0 0x0000779cc484321a in __GI___sigtimedwait (set=set@entry=0x7ffcd83d45f0, info=info@entry=0x7ffcd83d40c0, timeout=timeout@entry=0x0) at ../sysdeps/unix/sysv/linux/sigtimedwait.c:61
#1 0x0000779cc48428ac in __GI___sigwait (set=0x7ffcd83d45f0, sig=0x7ffcd83d4460) at ../sysdeps/unix/sysv/linux/sigwait.c:28
#2 0x00005fc059a41a34 in Nebula::start (this=0x7ffcd83d4e40, bootstrap_only=false) at src/nebula/Nebula.cc:1216
#3 0x00005fc059a39022 in oned_main () at src/nebula/oned.cc:107
#4 0x00005fc059a3942c in main (argc=2, argv=0x7ffcd83d5798) at src/nebula/oned.cc:243
Hello, we also build own version, but we use packaging system. We build RPM packages using docker in gitlab pipeline. You can get SPEC file from src package
The problem was not using systemd=yes. As we added that, the binary was compatible.