didnt know about nssm.exe! thanks! its like systemd services!!!
2 Likes
Self encryption?
I thought it is made by clients.
And node just store already encrypted data. Or not?
1 Like
it must be all the loggingā¦ but I might be far off
or the network chatter needs nodes to use more RAM?
1 Like
Thatās my assumption.
I think network buffers may have unrestricted growth.
But it is better to know for sure than to guess.
2 Likes
I had to stop my 10 nodes a few hours earlier after running them for 5-6hrs+ straightā¦ the bandwidth was exceeding 500Mbps+ā¦ (not the max upper limit of bandwidth here) but potentially causing issues with other services (locally)ā¦ (hitting close to upper max with other local services). I didnāt have time to figure out the root causeā¦ so I killed the 10 nodes in an urgencyā¦ (oops). They may not have been to blame for other issues seen locally, but I decided to stop them while stabilizing other local issues, lol.
Hopefully, 10 down nodes out of the ~120+ didnāt bring down the network, 
. Next time, I will implement QoS on the traffic coming out of the local nodes so it doesnāt impact the other traffic/services outside of the safe nodes from my local network, allowing them to run as is within an environment with an established dynamic upper bounds for safe network traffic.
One may argue, safe nodes should have highest priority on network traffic from a given environment (inbound/outbound) when safe farmingā¦ but this is all in the early daysā¦ 
.
8 Likes
oh no -.-" i have to admit my 10 nodes killed their safenode-processes somewhen this night ā¦ i guess it might have been that i was a bit cheap and just chose the smalleds droplets on DO that just have 1GB of RAM 
ā¦
9 Likes
My 8 nodes are on a 16GB RAM droplet and its almost full last time i checked! But all running still
11 Likes
I hope this tesnet exceeds expectations. 
10 Likes
I guess your approach with multiple nodes per instance was the better one 
5 Likes
If RAM is heavily used what happens if someone has problems with bad RAM sticks? As most RAM like DDR 4 has no error correction. A question as a experienced user of bad RAM sticks that only got discovered serveral years down the road as windows registry got problems from time to time.
Luckily RAM with error correction should be standard on DDR 5.
2 Likes
Worst case is the node fails.
5 Likes
As Mark says, itās just RAM and just a node. Not all hardware will work all the time etc. but thatās cool and as it should be. Our goal is to run on as small devices as we can and that is easiest with small nodes.
This testnet used a MemoryStore as the RecordStore just for simplicity. The next one will use a disk based RecordStore (Pr in flight IIRC). As soon as we write to disk we need to check for disk based conflicts etc. and we want to avoid configurations from users where at all possible. i.e. sensible defaults, but allow advanced users to say select the disk to store chunks on.
14 Likes
But that does not sound optimal. Would it maybe be a good idea to recommend example test net users to run memtest 86?
2 Likes
Iāve been trying to join a node from AWS and not succeeding so I thought Iād check things were still working by uploading some more files with the computer that was able to yesterday. I think something might be wrong now because the upload command is just hanging and Iām getting this in the log:-
[2023-05-06T10:56:03.151252Z TRACE safenode::client::api] For target 6e1175(01101110).., get closest peers Err(NotEnoughPeers)
[2023-05-06T10:56:03.151215Z ERROR safenode::client::api] Failed to dial 12D3KooWPuaKwn7rFNn4oj1ABDpikVZ7scHs29ZNCHBveLHgCmak: DialError(Transport([("/ip4/178.128.166.245/udp/36572/quic-v1/p2p/12D3KooWPuaKwn7rFNn4oj1ABDpikVZ7scHs29ZNCHBveLHgCmak", Other(Custom { kind: Other, error: HandshakeTimedOut }))]))
Are we dead?
1 Like
Why? Whatās special about a node that a user should do this to avoid an incredibly rare possibility that it will crash due to a single bit corruption?
Itās more likely to fail for many other reasons such as power cuts, user error, other programs having bugs etc.
The network is designed expecting nodes to fail regularly so it wonāt be affected.
Scaring users unnecessarily, or making them think they have to do special stuff to run a node when they donāt will discourage some and so hinder adoption for no discernable benefit.
5 Likes
the log of Writing 8 bytes to .. during upload actually means the target file not find.
make sure the suffix is presented correctly.
the safest way is put your targets under a folder and upload that folder as a whole
10 Likes
I just reconnected my node, this time via cable, not wifi. Thereās a lot of these messages in the newest logs:
[2023-05-06T11:47:41.059158Z INFO safenode::network::event] Dialing 12D3Koo...
A few inbound requests , too, but much less than yesterday. And also Adding peer to routing table seems to be happening.
Maybe we have lost and/or filled to max a number of nodes?
Yes, I am getting that and also:-
[2023-05-06T11:56:11.236224Z WARN safenode::network] Not enough peers in the k-bucket to satisfy the request
[2023-05-06T11:56:11.236225Z TRACE safenode::client::api] For target 3a1690(00111010).., get closest peers Err(NotEnoughPeers)
1 Like