One of my 3 nodes recently ran out of space and the moment the satellite stopped sending uploads (because the node ran out of space), the download requests stopped to. Since then the node only gets audits and repair downloads but no regular downloads anymore.
All nodes are on the same IP and host.
NodeID: 1cYRm7virx4Z7xXvgG9SxMBfJqkEA2XDoy6bWXfxkf2vLsmWjV
So since all non-full nodes are still getting a constant download while the full node doesn’t get any downloads (but audits and repair downloads), I assume this is a bug in the satellite software.
As Alexey told before here, most of data today are made by testers.
Testers mostly download more reason files, therefor full node, not have them and therefor have less usege.
The incoming data is split between 3 nodes, so the chances of one of 3 nodes not receiving ANY downloads in multiple days while the other 2 nodes receive just as many downloads as before is 0%.
Also it would be an extremely unlikely coincidences that the downloads stop exactly at the moment the node runs out of space and the satellite doesn’t send any uploads anymore.
test rules. For each 2-3 uploads you receive 1 download.
if your node is full, test traffic will stop. it wont try to download old data.
Detailed explanation:
your chart also confirms my suspects that the test traffic is being distributed in a steady way. All my nodes located in different locations of Europe are receiving same daily amound of data (+/-5%).
Of course, this rule can be changed at any time.
That could be a possible explanation.
maybe what Vadim meant but I didn’t understand it that way.
If only recently uploaded data is download then it makes sense the egress stops when ingress stops. But would also mean that my node now has 900GB of completely worthless testdata that is never going to be downloaded.
Our recent testing cycles involved uploading pieces to nodes that still have space and then redownloading them. If your node it already full, you will not get download traffic from those tests. You may get some egress from our first real customers if your node is holding some of their data and they decide they need to download some files. Your full node will still get paid for storing the pieces it holds, so this data is not completely worthless. We should be doing other types of testing soon to also download pieces from already full nodes. Real customers will probably have different usage patterns, but we still don’t expect them to generate huge amounts of download traffic at least in the beginning when they are just exploring the functionalities of the network. As we get closer to production, I would expect that our test data will start getting deleted slowly from full nodes to make room for real customer data as demand increases. Again, I would not label the test data as worthless, we do not expect higher download rates from real customer data at production release than what our tests are currently providing.
same would be the case if you had a node full of real customer data. You can add more disks on other drives, as you already have started doing according to your prior comments, if you want to increase capacity. If on the same IP, the effect will be the same as if your first disk had been bigger from the start. You will get more test data and more downloads on the still not filled diskspace. But remember, these tests will not go on forever especially after production release, so please don’t build a long term strategy based on getting huge amounts of egress from our tests. We hope that later on we will have customers that once they tested basic functionality, will start using Tardigrade for streaming, creating more egress revenue for our SNOs. But again, there are no guarantees when this may happen.
Of course, I was just worried that there could be a bug in the satellite software but since there is currently just this specific testing scenario and (almost) only testing data, all is good.
I expect the ingress and egress to become a lot lower and wildly fluctuating once tests are over and only customer data is stored.
the traffic in both directions stops with a available disk of 100GB. In reality i have only 57GB free, so i realized that my calculations for free space was wrong.
Now i am waiting for round about 3 days and nothing happens since that.
As nodes have now most only test data and testers upload and download only fresh data, then for now you not have uploads, this will change after all tests will be wiped and client data will moove
What I really do not understand are the following statements:
*) tardigrade has 1000s devs on the waitlist
*) storj traffic is really low - means no invitation / no dev testing
*) full nodes are full of test data, no uploads
So, can we hope for a bit more transparency where this is going?
dev invitations slow means that there is a technical problem?
we are currently onboarding the tardigrade customers (mostly devs who want to build apps on top of our platform) at a measured rate, and there are still thousands on the waitlist that have not been onboarded yet but will so soon. I myself do not have stats on how many have already been onboarded atm
Those customers who did get onboarded currently have a limit of 25GB/month uploads, so until we are ready to lift this restriction (when we are ready to charge them for the storage and bandwidth they use) they won’t be able to upload more than that, and in any case, we cannot force them to up or download files, they will do so as needed for their test purposes.
we have had other SNOs recently report receiving higher traffic than before, so you can’t generalize to the entire network from your own node’s traffic
As I already noted on other threads here, our test schemes are not constant, so you also cannot assume that full nodes will not receive traffic when our devs change the testing pattern to also test downloading from full nodes. It is correct that a full node will not get uploads once full, because there is no more room for more data.
As we also have stated before, we cannot onboard everyone all at once, we have to scale up carefully without exceeding our network’s capacity. And that is not just depending on how much space is being shared by SNOs.
I am not sure how we can be more transparent, I invite you to check other threads were we repeatedly have informed about our testing and customer onboarding, as well as making blog posts to further inform the public.
@heunland thank you so much for sharing this valuable info, very transparent!
And I totally agree that space is only one of the issues that need careful thought.
Keep up the good work, I am a happy SNO, this project is so exciting! I look forward to hearing more in the upcoming town hall Q1 2020.
