NodeDiscoveryNet [07/07/23 Testnet] [Maidsafe nodes offline]

[Profess]

NodeDiscoveryNet [07/07/23 Testnet] [Maidsafe nodes offline]

you didnt upload anything, thats just the example command line. you need / could upload a file. thats what it complains about. specify a real filename (pathname with filename) instead of the < path > you left there from the example command line

@loziniak

That is because you have to type a path to the file you want to upload instead <path> fragment.

Clear, thank you

[Profess]

@chriso

Hmm, it looks like you’ve got some kind of strange kind of setup going on with this computer.

None of the safe-related binaries require admin access.

I do not need to run PowerShell in administrator mode, it also works without it.

On the other hand, it was only when I started PowerShell ISE in administrator mode that the message appeared:
safeup.exe - system error - due to missing VCRUNTIME140.dll

the message did not appear before.

[Profess]

But I think I have finally got it right

[chriso]

@Profess

On the other hand, it was only when I started PowerShell ISE in administrator mode that the message appeared:
safeup.exe - system error - due to missing VCRUNTIME140.dll

Ah I see. That’s a useful piece of information - thanks!

Can I ask, which version of Windows are you running?

[Profess]

Glad I could be of help I use Windows 10 Pro

[chriso]

We had an install script before where installing the VC++ runtime was part of it. Might need to re-consider that again.

[Profess]

It would be good if someone else tried installing Testnet on Windows to see if they had a similar problem with a missing file.

But you should probably to change this command in the manual - the term ‘export’ is not recognized as the name of a cmdlet, function, script file, or operable program

@maidsafe

Using the Client

Once you have the client, you need to either set the SAFE_PEERS environmental variable or use the --peer= argument with any of the above network addresses.

Now to upload a directory/file to the network, use the following command:

# using the SAFE_PEERS variable (use 'SET' instead of 'export' on Windows) **[should be $env:SAFE_PEERS.. If set should be lower case I think - JL]**
export SAFE_PEERS="/ip4/206.189.21.247/tcp/45415/p2p/12D3KooWJa8AZnpp8uxUHHU7VJj1Stypm1Nt5bV3Yc62KnatRVFK"
safe files upload -- <path>

[chriso]

Yeah, the instructions for this particular testnet were inadequate.

To be honest, there is no real need to keep changing these. We should just have some boilerplate text we can copy and paste. Will make sure that’s sorted for subsequent testnets.

What you were missing on your machine was the VC++ runtime. I already had that installed on my machine, which is why it worked straight away for me. However, this is a dependency we’re already aware of.

[JPL]

How can I find the multiaddresses of my own nodes?

(for example "/ip4/165.22.123.138/tcp/41377/p2p/12D3KooWBPdhoD9uSbtwFeiQWmTZT2Lhaz2HRrgtLWauMwYQK1pK"

Edit: found it

sn_networking::event] Local node is listening on "/ip4/176.58.124.107/tcp/37421/p2p/12D3KooWEFRkqnaJ3Dzpnq3qWu1DbQc3uqHC9GehqpBKdLYg9oZr"

[storage_guy]

I was about to reply but you got it! Just grep for ‘listening’ in the first log file created.

The awkward thing is that this message seems to only appear once and when the node starts up. So when the log file is eventually deleted after about a week you can never tell what your node id is! You can find your port number and you should know your IP addresses but not the node id.

Feature request: have the ‘listening’ string repeated in the logs every few mins.

[anon26713768]

Perhaps saved in a file same as pid is?

[Shu]

I found it useful using the safenode_rpc_client if you start up your node with --rpc 127.0.0.1:1234:

Then safenode_rpc_client 127.0.0.1:1234 netinfo shows:

...
Node's listeners:
Listener: /ip4/127.0.0.1/tcp/12000
Listener: /ip4/192.168.X.Y/tcp/12000

In addition doing safenode_rpc_client 127.0.0.1:1234 info shows:

Node info:
==========
RPC endpoint: https://127.0.0.1:1234
Peer Id: <YourPeerId>
Logs dir: <YourLogPathDir>
PID: #
Binary version: 0.86.5
Time since last restart: 253094s

[Shu]

@joshuef

We’ve just been looking over some logs and found a suspiciously high amount of connections going to each peer (sometimes even hundreds!). This is something libp2p should be handling for us, so either we’re doing something wrong , or we’ve found a bug!.

I too am seeing up to 100s of connected peers per node (instant report) when scanning for connected peers on a per minute basis via the RPC method noted here. Below is an updated graph from the original comments noted here :

Updates:

• Added panel for # Unique Peers Connected per Min

[joshuef]

We’ve been digging deeper here this week and it seems:

• there’s 2 conns for every peer. And kad should maintain conns for all peers in the routing table. This would mean ~150 open conns on each node would make sense
• multiplexing should mean we use the same port for all IN conns and all OUT.
• Right now we open alot of ports, which seems to be because of a libp2p default dial condition meaning the kad layer is opening hundreds of connections on new ports.
• this is memory intense.

There’s fixes for that and other conn related things in the pipeline

[joshuef]

An update on the testnet in general:

Mem issues are obviously still about in this branch, and we’re down to 11 nodes/droplet now.

With that, we’ve only got two “full” nodes so far (and those could still be pruning data).

mem on the surviving nodes is ~350mb atm.

I can still get our test data fine

[Toivo]

@joshuef

There’s fixes for that and other conn related things in the pipeline

Your pipeline or libp2p’s?

[dirvine]

Both really. Libp2p needs to get QUIC as first class hole punch citizen and we need to work with what they have and we can contribute to. So nothing to worry us, but progress needs to happen.

While that happens we can move forward to finalise upgrades, payments etc. So all good

[joshuef]

Aye, @bzee has worked up some PRs for libp2p which are waiting in the wings there.

We have that tested but can’t merge until a release there.

And then we’ve other related work in the safe codebase on the go too.

[anon26713768]

@joshuef

we’re down to 11 nodes/droplet

Lost 89% of the original nodes and the test data is still there.
Damn fine work gentlemen

[dirvine]

When this is truly multiplexed (QUIC) then I feel hundreds of nodes per PC is gonna be simple. Great to see kad doing what kad should do here and doing it well. @joshuef does a bunch of churn tests with 6 times 100% node loss and it’s very good, really good.