mirror of
https://github.com/haveno-dex/haveno.git
synced 2024-11-16 15:58:08 +00:00
246 lines
9.4 KiB
Markdown
246 lines
9.4 KiB
Markdown
# Haveno Network Monitor Node
|
|
|
|
The Haveno monitor node collects a set of metrics which are of interest to developers and users alike. These metrics are then made available through reporters.
|
|
|
|
The *Settled* release features these metrics:
|
|
- Tor Startup Time: The time it takes to start Tor starting at a clean system, unpacking the shipped Tor binaries, firing up Tor until Tor is connected to the Tor network and ready to use.
|
|
- Tor Roundtrip Time: Given a bootstrapped Tor, the roundtrip time of connecting to a hidden service is measured.
|
|
- Tor Hidden Service Startup Time: Given a bootstrapped Tor, the time it takes to create and announce a freshly created hidden service.
|
|
- P2P Round Trip Time: A metric hitchhiking the Ping/Pong messages of the Keep-Alive-Mechanism to determine the Round Trip Time when issuing a Ping to a seed node.
|
|
- P2P Seed Node Message Snapshot: Get absolute number and constellation of messages a fresh Haveno client will get on startup. Also reports diffs between seed nodes on a per-message-type basis.
|
|
- P2P Network Load: listens to the P2P network and its broadcast messages. Reports every X seconds.
|
|
- P2P Market Statistics: a demonstration metric which extracts market information from broadcast messages. This demo implementation reports the number of open offers per market.
|
|
|
|
|
|
The *Settled* release features these reporters:
|
|
- A reporter that simply writes the findings to `System.err`
|
|
- A reporter that reports the findings to a Graphite/Carbon instance using the [plaintext protocol](https://graphite.readthedocs.io/en/latest/feeding-carbon.html#the-plaintext-protocol)
|
|
|
|
## Configuration
|
|
|
|
The *Haveno Network Monitor Node* is to be configured via a Java properties file. There is a default configuration file shipped with the monitor which reports to the one monitoring service currently up and running.
|
|
|
|
If you want to tweak the configuration, you can pass the location of the file as command line parameter:
|
|
|
|
```
|
|
./haveno-monitor /path/to/your/config.properties
|
|
```
|
|
|
|
A sample configuration file looks like follows:
|
|
|
|
```
|
|
## System configuration
|
|
|
|
# true overwrites the reporters picked by the developers (for debugging for example) (defaults to false)
|
|
System.useConsoleReporter=true
|
|
|
|
# 0 -> XMR_MAINNET, 1 -> XMR_LOCAL (default)
|
|
System.baseCurrencyNetwork=0
|
|
|
|
## Each Metric is configured via a set of properties.
|
|
##
|
|
## The minimal set of properties required to run a Metric is:
|
|
##
|
|
## YourMetricName.enabled=true|false
|
|
## YourMetricName.run.interval=10 [seconds]
|
|
|
|
#Edit and uncomment the lines below to your liking
|
|
|
|
#TorStartupTime Metric
|
|
TorStartupTime.enabled=true
|
|
TorStartupTime.run.interval=100
|
|
TorStartupTime.run.socksPort=90500 # so that there is no interference with a system Tor
|
|
|
|
#TorRoundTripTime Metric
|
|
TorRoundTripTime.enabled=true
|
|
TorRoundTripTime.run.interval=100
|
|
TorRoundTripTime.run.sampleSize=5
|
|
TorRoundTripTime.run.hosts=http://expyuzz4wqqyqhjn.onion:80 # torproject.org hidden service
|
|
|
|
#TorHiddenServiceStartupTime Metric
|
|
TorHiddenServiceStartupTime.enabled=true
|
|
TorHiddenServiceStartupTime.run.interval=100
|
|
TorHiddenServiceStartupTime.run.localPort=90501 # so that there is no interference with a system Tor
|
|
TorHiddenServiceStartupTime.run.servicePort=90511 # so that there is no interference with a system Tor
|
|
|
|
#P2PRoundTripTime Metric
|
|
P2PRoundTripTime.enabled=true
|
|
P2PRoundTripTime.run.interval=100
|
|
P2PRoundTripTime.run.sampleSize=5
|
|
P2PRoundTripTime.run.hosts=723ljisnynbtdohi.onion:8000, fl3mmribyxgrv63c.onion:8000
|
|
P2PRoundTripTime.run.torProxyPort=9060
|
|
|
|
#P2PNetworkLoad Metric
|
|
P2PNetworkLoad.enabled=true
|
|
P2PNetworkLoad.run.interval=100
|
|
P2PNetworkLoad.run.torProxyPort=9061
|
|
P2PNetworkLoad.run.historySize=500
|
|
|
|
#P2PNetworkMessageSnapshot Metric
|
|
P2PSeedNodeSnapshot.enabled=true
|
|
P2PSeedNodeSnapshot.run.interval=24
|
|
P2PSeedNodeSnapshot.run.hosts=3f3cu2yw7u457ztq.onion:8000, 723ljisnynbtdohi.onion:8000, fl3mmribyxgrv63c.onion:8000
|
|
P2PSeedNodeSnapshot.run.torProxyPort=9062
|
|
|
|
#P2PMarketStats Metric
|
|
P2PMarketStats.enabled=true
|
|
P2PMarketStats.run.interval=37
|
|
P2PMarketStats.run.hosts=ef5qnzx6znifo3df.onion:8000
|
|
P2PMarketStats.run.torProxyPort=9063
|
|
|
|
#PriceNodeStats Metric
|
|
PriceNodeStats.enabled=true
|
|
PriceNodeStats.run.interval=42
|
|
PriceNodeStats.run.hosts=http://5bmpx76qllutpcyp.onion, http://xc3nh4juf2hshy7e.onion, http://44mgyoe2b6oqiytt.onion, http://62nvujg5iou3vu3i.onion, http://ceaanhbvluug4we6.onion
|
|
|
|
#MarketStats Metric
|
|
MarketStats.enabled=true
|
|
MarketStats.run.interval=191
|
|
|
|
## Reporters are configured via a set of properties as well.
|
|
##
|
|
## In contrast to Metrics, Reporters do not have a minimal set of properties.
|
|
|
|
#GraphiteReporter
|
|
GraphiteReporter.serviceUrl=k6evlhg44acpchtc.onion:2003
|
|
|
|
```
|
|
|
|
## Run
|
|
|
|
The distribution ships with a systemd .service file. Validate/change the executable/config paths within the shipped `haveno-monitor.service` file and copy/move the file to your systemd directory (something along `/usr/lib/systemd/system/`). Now you can control your *Monitor Node* via the usual systemd start/stop commands
|
|
|
|
```
|
|
systemctl start haveno-monitor.service
|
|
systemctl stop haveno-monitor.service
|
|
```
|
|
and
|
|
```
|
|
systemctl enable haveno-monitor.service
|
|
```
|
|
|
|
You can reload the configuration without restarting the service by using
|
|
|
|
```
|
|
systemctl reload haveno-monitor.service
|
|
```
|
|
|
|
Follow the logs created by the service by inspecting
|
|
|
|
```
|
|
journalctl --unit haveno-monitor --follow
|
|
```
|
|
|
|
# Monitoring Service
|
|
|
|
A typical monitoring service consists of a [Graphite](https://graphiteapp.org/) and a [Grafana](https://grafana.com/) instance.
|
|
Both are available via Docker-containers.
|
|
|
|
## Setting up Graphite
|
|
|
|
### Install
|
|
|
|
For a docker setup, use
|
|
|
|
```
|
|
docker run -d --name graphite --restart=always -p 2003:2003 -p 8080:8080 graphiteapp/graphite-statsd
|
|
```
|
|
|
|
- Port 2003 is used for the [plaintext protocol](https://graphite.readthedocs.io/en/latest/feeding-carbon.html#the-plaintext-protocol) mentioned above
|
|
- Port 8080 offers an API for user interfaces.
|
|
|
|
more information can be found [here](https://graphite.readthedocs.io/en/latest/install.html)
|
|
|
|
### Configuration
|
|
|
|
For configuration, you must adapt the whisper database schema to suit your needs. First, stop your docker container by running
|
|
|
|
```
|
|
docker stop graphite
|
|
```
|
|
|
|
Find your config files within the `Source` directory stated in
|
|
|
|
```
|
|
docker inspect graphite | grep -C 2 graphite/conf\",
|
|
```
|
|
|
|
Edit `storage-schemas.conf` so that the frequency of your incoming data (configured in the monitor configs `interval`) is matched. For example, insert
|
|
```
|
|
[haveno]
|
|
pattern = ^haveno.*
|
|
retentions = 10s:1h,5m:31d,30m:2y,1h:5y
|
|
```
|
|
before the `[default...` blocks of the file. This basically says, that every incoming set of data reflects 5 minutes of the time series. Furthermore, every 30 minutes, the data is compressed and thus, takes less memory as it is kept for 2 years.
|
|
|
|
Further, edit `storage-aggregation.conf` to configure how your data is compressed. For example, insert
|
|
```
|
|
[haveno]
|
|
pattern=^haveno.*
|
|
xFilesFactor = 0
|
|
aggregationMethod = average
|
|
```
|
|
before the `[default...` blocks of the file. With this configuration, whenever data is aggregated, the `average` data is made available given that at least `0%` of the data points (i.e. floor(30 / 5 * 40%) = 2 data points) exist. Otherwise, the aggregated data is dropped. Since we start the first hour with a frequency of 10s but only supply data every 4 to 6 minutes, our aggregated values would get dropped.
|
|
|
|
*Please note, that I have not been able to get the whole thing to work without the 10s:1h part yet*
|
|
|
|
Finally, update the database. For doing that, go to the storage directory of graphite, the `Source` directory stated in
|
|
```
|
|
docker inspect graphite | grep -C 2 graphite/conf\",
|
|
```
|
|
Once there, you have two options:
|
|
- delete the whisper directory
|
|
```
|
|
rm -r whisper
|
|
```
|
|
- update the database by doing
|
|
```
|
|
find ./ -type f -name '*.wsp' -exec whisper-resize.py --nobackup {} 10s:1h 5m:31d 30m:2y 1h:5y \;
|
|
```
|
|
|
|
and finally, restart your graphite container:
|
|
```
|
|
docker start graphite
|
|
```
|
|
|
|
|
|
Other than that, there is no further configuration necessary. However, you might change your iptables/firewalls to not let anyone access your Graphite instance from the outside.
|
|
|
|
### Backup your data
|
|
|
|
The metric data is kept in the `Source` directory stated in
|
|
```
|
|
docker inspect graphite | grep -C 2 graphite/conf\",
|
|
```
|
|
ready to be backed up regularly.
|
|
|
|
## Setting up Grafana
|
|
|
|
### Install
|
|
|
|
For a docker setup, use
|
|
|
|
```
|
|
docker run -d --name=grafana -p 3000:3000 grafana/grafana
|
|
```
|
|
|
|
- Port 3000 offers the web interface
|
|
|
|
more information can be found [here](https://grafana.com/grafana/download?platform=docker)
|
|
|
|
### Configuration
|
|
|
|
- Once you have Grafana up and running, go to the *Data Source* configuration tab.
|
|
- Once there click *Add data source* and select *Graphite*.
|
|
- In the HTTP section enter the IP address of your graphite docker container and the port `8080` (as we have configured before). E.g. `http://172.170.1:8080`
|
|
- Select `Server (default)` as an *Access* method and hit *Save & Test*.
|
|
|
|
You should be all set. You can now proceed to add Dashboards, Panels and finally display the prettiest Graphs you can think of.
|
|
A working connection to Graphite should let you add your data series in a *Graph*s *Metrics* tab in a pretty intuitive way.
|
|
|
|
- Optional: hide your Grafana instance behind a reverse proxy like nginx and add some TLS.
|
|
- Optional: make your Grafana instance accessible via a Tor hidden service.
|
|
|
|
### Backup your data
|
|
|
|
Grafana stores every dashboard as a JSON model. This model can be accessed (copied/restored) within the dashboard's settings and its *JSON Model* tab. Do with the data whatever you want.
|