nullTracers

[coot]

Checking the cardano-explorer code I found a few nullTracers which should not be in a production code. This will not allow us to debug the application, and should not be in a production code.

• ncSubscription_V1 traces
• handshake tracer
• mux tracer
• chain-sync tracer

[erikd]

Is there a way to have only Error level messages come out of the network, but have Debug for explorer?

[coot]

Sorry, it seems I missed your message; I think there is. cardano-node has an extensive configuration of traces and debug levels, check this out: mux-tracer and handshake levels

[erikd]

I have looked at the ToObjectOrphans.hs module in cardano-node (it has moved to https://github.com/input-output-hk/cardano-node/blob/master/cardano-config/src/Cardano/Tracing/ToObjectOrphans.hs ).

That is 1300 lines of code which I think is rather a lot. That gives me two options, copy/paste it from node which seems insane, or pull in cardano-node as a dependency (db-sync does not otherwise depend on node (which also seem insane).

Even if I do one of the above, what do I replace the nullTracer with?

[coot]

Yes, copying all is not good, a long term solution, in my opinion, is to add cardano-monitoring-framework library which has all the instantiations, and is easily reusable by other clients.

Even if I do one of the above, what do I replace the nullTracer with?
You must have a top level tracer. It should be general enough to use its contravarince to transform it to the form required by the network tracers. cardano-node does this in Cardano.Tracing.Tracers. Again there's a lot of code, some more specific to the node (e.g. consensus tracers, network tracers for node to node protocols).

There's also work on cli tools, but it lives inside cardano-node. They should pull it out and set a proper boundaries between clients and the node code.

[disassembler]

The current problem we don't know how to debug is the log shows:

Starting cardano-db-sync.service...
Connection to localhost 5432 port [tcp/postgresql] succeeded!
Started cardano-db-sync.service.
[db-sync-node:Info:1] [2020-03-04 15:46:34.46 UTC] NetworkMagic: RequiresMagic 1097911065
[db-sync-node:Info:1] [2020-03-04 15:46:34.46 UTC] Initial genesis distribution present and correct
[db-sync-node:Info:1] [2020-03-04 15:46:34.46 UTC] Total genesis supply of Ada: 41999999999.999903
[db-sync-node:Info:1] [2020-03-04 15:46:34.47 UTC] localInitiatorNetworkApplication: connecting to node via "/run/cardano-node/node.socket"

if we strace the process we see occurring over and over again:

bind(4, {sa_family=AF_UNIX, sun_path="/run/cardano-node/node.socket"}, 110) = -1 EADDRINUSE (Address already in use)
socket(AF_UNIX, SOCK_STREAM|SOCK_NONBLOCK, 0) = 4

If we can get logs working, we should be able to identify what the actual problem is.

[erikd]

Proper tracers are in #8 which has not yet been merged due to CI issues.

[coot]

Still with the proper tracers we couldn't see anything useful in the logs, even if enabled Debug severity (see my branch coot/haskell-updates. @karknu replaced the logging trace with stdoutTracer and was able to get the information that we needed to debug the connection issue. This suggest that there's something wrong with the configuration of the loggers.

^ @CodiePP

[erikd]

The tracers fix has been merged to master. Closing this.