codechain-io / codechain-dashboard Goto Github PK

Currently, the server sends an email and slack message only when there is a problem.
So the administrator is hard to know whether the server is alive or the server is silently gone.
Let's make it send a notification daily.

I don't think about the content of the notification, but I want the title of the email is the following format.

[info][{network id}][dashboard-server] is working - {Utc::now().to_rfc3339()}

Current dashboard just reports whether the node is 'Running' or 'Stopped'. Sometimes it isn't sufficient to diagnosis the error.

We should report when the node's best block number is far behind or spurious errors are flooding.

We renamed agent -> dashboard-client and agent-server(agent-hub) -> dashboard-server. The domains are left as un changed. Let's change them.

Also, change the Corgi's URL: agent-server.corgi.codechain.io

Ubuntu automatically updates PostgreSQL for the security. The dashboard server should reconnect when PostgreSQL is restarted. There are two parts that are using PostgreSQL connection. One is db_service and the other one is remove_network_usage. db_service is already using the connection pool but remove_network_usage is not using the connection pool.

We may use the pg_cron

The agent(dashboard client) recognizes ipc path in the command line argument, but it doesn't understand the ipc path specified in the config file.

Currently, the agent server only checks total usable disk space. However, if a server has more than one volume, the agent server should send alarms for each volume.

Currently, we are updating the node by compiling the new version of CodeChain in every node. However, the node may not have enough hardware spec for building CodeChain.

It is better to support updating CodeChain by downloading a binary file from the given URL.

We are removing rows in the network_usage table that were inserted more than a week ago to reduce the usage of disk. We use the network usage table to find strange network throughputs. Since we will check the strange network behavior when it appeared, we don't need to save the previous data.

The purpose of the peer_count table is the same as the network_usage table. We need to remove old rows in the peer_count table.

The Agent Hub uses too many threads to running concurrently.

The information this script shows needs to be added on dashboard.

PostgreSQL could be restarted by an automatic security update.

See https://help.ubuntu.com/community/AutomaticSecurityUpdates

We are using spot instances in the network.
Spot instances could be shut down anytime.

The agent should start automatically when the spot instance is started.

Current dashboard server does not support PostgreSQL 12. If you run generate-schema it fails with the 12 version.

I set the JSON-RPC port to 18080 in the config file, but agent-hub sends pings to 8080.

In some cases, the Agent Server is using the NETWORK_ID from agents. Since the NETWORK_ID could be empty, using the NETWORK_ID from environment variable is better.

I tried to install server but generate-schema failed with the following messages.

     Running `target/debug/generate-schema`
#2019-06-05 13:39:13  main INFO agent-hub  Create agent_extra table
#2019-06-05 13:39:13  main INFO agent-hub  Create logs table
#2019-06-05 13:39:13  main INFO agent-hub  Create logs_timestamp index
#2019-06-05 13:39:13  main INFO agent-hub  Create logs_target index
#2019-06-05 13:39:13  main INFO agent-hub  Create network_usage table
#2019-06-05 13:39:13  main INFO agent-hub  Create network_usage_time_index
#2019-06-05 13:39:13  main INFO agent-hub  Create materialized views
thread 'main' panicked at 'called `Result::unwrap()` on an `Err` value: Error(Db(DbError { severity: "ERROR", parsed_severity: Some(Error), code: SqlState("42P01"), message: "relation \"peer_count\" does not exist", detail: None, hint: None, position: Some(Normal(259)), where_: None, schema: None, table: None, column: None, datatype: None, constraint: None, file: Some("parse_relation.c"), line: Some(1180), routine: Some("parserOpenTable") }))', src/libcore/result.rs:997:5
note: Run with `RUST_BACKTRACE=1` environment variable to display a backtrace.

I used PostgreSQL 10.8.

psql (PostgreSQL) 10.8 (Ubuntu 10.8-0ubuntu0.18.04.1)

network_usage table takes too much disk space.

Currently, the client thinks the port of the node is 3485 if there is no argument.

The dashboard just got one connection removed action though there were four connections more.

We could provide a Docker file which set up environments.

We may use Docker Compose.
The cluster may have 1 of the Agent Hub and 2 of Agents.

First page

1. Network Out bytes of this week(All nodes)
2. (Network Out bytes / peer count) of this week(All nodes)
3. Network Out bytes of this week(To Known Nodes, To Unknown Nodes)
   

Detail page

If you click a machine's name in the first page, you can see the detail page.

1. Network Out bytes of this week(Per network extension)
2. Network Out bytes of this week(Per to IP address)
   

Because the content of the modal is not scrollable, the user cannot see the whole list if they have a small display.

I ran the dashboard and I show the following error.
It fixed when I restarted the dashboard server.
Please check it.

The Dashboard already has a UI to send JSON RPC call to CodeChain nodes.
We should implement API for the Dashboard.

Currently, the nodes list is shown in the fixed order.
How about changing it to be sortable by name/IP/block number/version/commit hash?

It is very common to stop and start all nodes after a few modifications.

Please make it vertically resizable.

The number of connection is also important for stable block generation.
But there is no way to check it easily for now.
How about showing the number of established connections in the node list and make it sortable.

CodeChain agent tries to find & update the .git folder in the CodeChain directory, and build a new binary.
However, most of the machines running CodeChain don't have enough performance to build a binary, so it's useless now.
We might need some kind of auto-updating feature for CodeChain agent in the future, but it will be different from the current implementation.