Daily Qntra production continued as I've been wrestling with the NSD name server daemon. With two of the four servers still getting loaded with blockchain, I brought the nameserving orchestra online New Years Eve. Five boxes, the 'anyserver' hosting Qntra alongside a number of aligned blogs was joined by four other servers spread around doing the Bitcoin node thing and other tasks to be determined were all gloriously serving Qntra DNS for two days. Then they stopped, and reversion to the prior setup that didn't shit the bed was done.
NSD was chosen over other DNS serving daemons like BIND and Unbound because it does one thing well. It answers port 53 with authority on the zones it servers DNS info for. NSD doesn't recurse nor does it resolve. It does however contain a mechanism to have several nameservers serving the same zones sync through what the NSD developers describe as a "master-slave" arrangement. I'd picked one of the five boxes to do the mastering for the others, and...
NSD shat the bed over the servers disagreeing on what time it is producing numerous loglines of the:
... query: bad tsig (Bad Time) for key sec_key. ...
variety. Now, importing the empire's political time and running a daemon on all of the servers just to fuck the clocks into agreeing for the sake of agreeing is out of the question. Reading the docs from NSD's developers at "Stichting NLnet Labs" hasn't produced a good way to get the NSD servers to talk to each other that doesn't import things or leave the clock drift time bomb ticking. Thusly I'll be bringing the orchestra back up without any means for the NSD instances to talk to each other and just keeping them in sync myself. There aren't that many names being served at present.
All four on the servers have managed to catch up from the blockchain snapshots they were loaded with and as of this writing, they are keeping pace with the network.
The Servers With TRB Nodes
- 205.134.172.4 - This node lives in Alf's Rack and advertises a 70001 version number. One reason for not flying the 99999 flag on this node is that one of the neighbors is, and if they don't another will. The other reason is that PRB nodes have taken to using their segwit shit to pass around blocks bigger than 1MB with the segwit shit included. TRB won't eat these. Having some nodes with version strings under the Power Ranger retard level should assist in the propagation of blocks without the SegWit imaginarium glued on.
- 88.80.148.58 - Bulgaria
- 192.151.158.26 - Kansas, same low bidder operation 'anyserver' is provided by.
- 143.202.160.10 - Costa Rica
TRB nodes reject obese blocks regardless. But setting e.g. 70001 does save bandwidth, as PRB peers will abort prior to taking a multi-MB shit into the pipe.
Rejecting bad fakeblocks is good, but the actual blocks still have to be had.