Closed bobbyvinon closed 7 years ago
here is backtrace:
at osdep/Thread.hpp:114
Thread 1 (Thread 0x7ff6fcb68740 (LWP 4304)):
........
node=<optimized out>, dbPath=0x13e9730 "/var/lib/zerotier-one/controller.d")
at controller/EmbeddedNetworkController.cpp:434
at service/OneService.cpp:752
I had the same issue. Didn't have time to debug, reverting to 1.2.2 fixed it.
i've attached config files, which json failed to parse a.zip
dirty quick fix: in function JSONDB::get comment this lines at beginning of function: while (!_ready) { Thread::sleep(250); _ready = _reload(_basePath,std::string()); }
I have the same issue.
I also have the same issue for a long time now, will double confirm if the quickfix by @bobbyvinon works later on.
git bisect says
first bad commit: [f4feccc6265cc480b84c85f897b225714072d4ec] Do not serve controller requests until init is done.
'bad' meaning one
hangs at start and has to be kill -9'd
I was testing with this in controller.d/network/54d343738795f1c8.json
{
"authTokens": [],
"capabilities": [],
"creationTime": 1509777679437,
"enableBroadcast": true,
"id": "54d343738795f1c8",
"ipAssignmentPools": [],
"lastModified": 1509777679436,
"multicastLimit": 32,
"name": "",
"nwid": "54d343738795f1c8",
"objtype": "network",
"private": true,
"revision": 1,
"routes": [],
"rules": [
{
"not": false,
"or": false,
"type": "ACTION_ACCEPT"
}
],
"tags": [],
"v4AssignMode": {
"zt": false
},
"v6AssignMode": {
"6plane": false,
"rfc4193": false,
"zt": false
}
}
which was generated by starting a controller, adding a network, exiting. ( I was testing something in node)
I can confirm similar behaviour to @laduke. I thought I was getting configurations wrong because joined users would not receive IPv4 addresses. (Re)starting the service again results in a hang that must be kill -9
'd. However, using the most basic case, where I start the service and add a network with all defaults, then restarting the service causes the hang.
So, I don't think it's anything I've configured incorrectly. How is anyone using their own controller at this state? There should be more documentation around the routes
configuration, because it's necessary to setup addressing but the payload had to be guessed by reading around non-tierzero sites (which there's very little information for in any case).
Same here. Doing a git revert f4feccc6
fixes the problem.
Fixed since that code got taken out behind the barn and shot.
I've built latest zerotier from github (ubuntu 16.04 with clang). It's unmodified build with default config. Only config i changed is local.conf:
{ "settings": { "primaryPort": 9991, "allowManagementFrom": [ "0.0.0.0/0" ] } }
zerotier worked some time. i've attached two clients to this controller, all worked fine. after controller server reboot it stopped working. it starts, receives incoming connections but didn't answers to them - just hang. i've removed /var/lib/zerotier and reinstalled package again. situation are the same - it worked some time and after few reboots stopped to work.