nspcc-dev / coredns Goto Github PK

We should create separate neofs-health plugin (that uses healht logic from geodns plugin).

One NNS in mainnet (higher priority) and one in sidechain. Merge result.

Support SRV record in NNS plugin after nspcc-dev/neofs-contract#254

The plugin should allow to set the following params:

• Health check method
• Health check request timeout
• Cache size
• Cache expiration and health check interval (initiate health status update when the cache entry expires)
• Filter (check records name for full matching), must be the last because there can be multiple names for filtering

Workflow:

1. Get records
2. If records doesn't match filter -> return all records
3. If records match filter -> return health records from cache (if cache is empty return all records) and trigger health check (add records to pool that is checked periodically to update health status)

As @532910 noticed healthcheck plugin sends to many requests.
We should use a different key format for caching. Currently, it may happen that cache contains two different records with the same IP to check.

What happened:

sergio@lily:~$ host cdn.fs.neo.org
cdn.fs.neo.org has address 207.154.194.79
sergio@kiki:~$ host cdn.fs.neo.org
cdn.fs.neo.org has address 207.154.194.79

kiki is in Frankfurt, lily is in Singapore. On both requests DNS server responds with kiki's IP.

What you expected to happen:
When I request cdn.fs.neo.org from these servers, I expect that the DNS responds with the server's IP (from kiki with 207.154.194.79, from lily with 167.71.206.254).

Environment:
The version of CoreDNS: b668b7f8981d67902080b5a961e7ae8a8ff2b461
Corefile

cdn.fs.neo.org:53 {
    whoami
    log
    debug
    file zones/cdn.fs.neo.org
    cache 3600
    health :8080
    geodns /var/lib/GeoIP/
    #nns https://rpc1.morph.fs.neo.org:40341

}

zones/cdn.fs.neo.org

$ORIGIN cdn.fs.neo.org.
$TTL 3600
@ 3600 IN       SOA ns1.fs.neo.org. ns2.fs.neo.org. 2017042745 7200 3600 1209600 3600
cdn.fs.neo.org.  IN  NS      ns1.fs.neo.org.
cdn.fs.neo.org.  IN  NS      ns2.fs.neo.org.
cdn.fs.neo.org.  IN  A       207.154.194.79
cdn.fs.neo.org.  IN  A       167.71.206.254
cdn.fs.neo.org.  IN  AAAA    fe80::8e8:7eff:fef1:1207
cdn.fs.neo.org.  IN  AAAA    fe80::8da:87ff:fe24:7d6b
www              IN  CNAME   cdn.fs.neo.org.

Logs on request to DNS server from office (92.255.97.106 is the office's public IP)

окт 22 15:50:14 moro coredns[284885]: [WARNING] plugin/geodns: couldn't get location 92.255.97.106 from db: not found
окт 22 15:50:14 moro coredns[284885]: [INFO] 92.255.97.106:39358 - 52001 "A IN cdn.fs.neo.org. udp 32 false 512" NOERROR qr,aa,rd 176 0.000197227s
окт 22 15:50:14 moro coredns[284885]: [WARNING] plugin/geodns: couldn't get location 92.255.97.106 from db: not found
окт 22 15:50:14 moro coredns[284885]: [INFO] 92.255.97.106:38475 - 14401 "AAAA IN cdn.fs.neo.org. udp 32 false 512" NOERROR qr,aa,rd 200 0.000261954s
окт 22 15:50:14 moro coredns[284885]: [DEBUG] plugin/geodns: unsupported type MX, nothing to do
окт 22 15:50:14 moro coredns[284885]: [INFO] 92.255.97.106:48568 - 21035 "MX IN cdn.fs.neo.org. udp 32 false 512" NOERROR qr,aa,rd 110 0.000141053s

окт 22 15:51:25 moro coredns[284885]: [WARNING] plugin/geodns: couldn't get location 199.247.0.222 from db: not found
окт 22 15:51:25 moro coredns[284885]: [INFO] 199.247.0.222:48404 - 51282 "A IN cdn.fs.neo.org. udp 55 false 4096" NOERROR qr,aa,rd 176 0.001178447s

After #20 add an other way to set filter for record names -- regexp.

Need to add a new param to NNS plugin config in order to change the domain of request (e.g. from containers.testnet.fs.neo.org to containers).

Example old config:

containers.testnet.fs.neo.org:53 {
    nns https://rpc1.morph.fs.neo.org:40341
}

Example new config:

containers.testnet.fs.neo.org:53 {
    nns https://rpc1.morph.fs.neo.org:40341 containers
}

So request nicename.containers.testnet.fs.neo.org to coredns will transform to request nicename.containers to NNS contract.

After #20 add other methods to check health (not only via http request, but i.e. icmp_ping)

What would you like to be added:
NNS plugin support.

We should use grpc first

We need to implement geodns functionlity.

First we should support at least (NNS compatible records) and determining exact min distance server (geodns support weighted servers and more 5% minimal distance)

What would you like to be added:
When edge server request fails on failed healthcheck, I'd like to see a log message of level WARN formatted like below:

tried to resolve the <hostname>, healthchecked <IP>:<port>, healthcheck failed, no answer returned

Why is this needed:
For better troubleshooting when deploying DNS server by operations team

Version which should be tested: 3ca07a4 (last commit of master branch)

How to configure: https://github.com/nspcc-dev/coredns/blob/master/plugin/healthchecker/README.md

What would you like to be added:

I'd like to have an ability to specify multiple GeoIP databases or a path to the directory with these databases.

Why is this needed:

There are many GeoIP databases (with cities, countries, ASNs, etc) and I need all of them in DNS server configuration.

Current implementation is targeted for sidechain where NNS contract is deployed with ID:1

As for NNS contracts in mainchain, we should support optional contract address configuration.

We should add transfer support (e.g. axfr request using dig). It can be done using built-in transfer plugin but it must be supported in nns contract first (so issue is blocked by this).

It would be nice to update neo-go to v0.99.3 and replace deprecated code. https://github.com/nspcc-dev/neo-go/releases/tag/v0.99.3
But probably there is some problem with go version 1.16 in go.mod