Ne00n

Could be that the memory can't be allocated inside the virtual machine because the host system has already ran out of memory. I would suggest you verify that your virtual...

I nailed it down to the config file, apparently gdnsd is not happy with a lot of resource maps. edit: looks more like it can be a combination between the...

I tried to reproduce it locally and it seems I figured the point when it goes into limbo. Basically, I took the same nets mapping file, about 1 million rows....

The magic number is indeed 13, if I go over 13 datacenters and 13 * 13 maps/resources it goes nuts.

Example files: https://raw.githubusercontent.com/Ne00n/gdns-configs/master/config.works (13 datacenters) https://raw.githubusercontent.com/Ne00n/gdns-configs/master/config.broken (14 datacenters)

Well, as I said before, I can't reload the config while the server is running, because the your decision was, to not support that. So my issue is, if I...

Well, its a single point of failure plus it does only measures from a single point. Which makes not an option, its just not reliable enough. Of course I would...

I uploaded the dc.conf file here if you wanna take a look: https://github.com/Ne00n/gdns-configs/blob/master/dc.conf Should match the .broken config

Thanks, I check tomorrow. Well, I use rqlite as database, which uses Raft Consensus Algorithm to make these decisions. rqlite is aware which nodes are available, I use that information...

Works so far and looks stable. Yeah the initial boot phase exceeds the default systemd timeout but can be extended. I don't restart it often anyway and have no reason...