Updates on Test Data

I can understand IOPS being a concern: but I think you’ll be fine for capacity. I get a strong sense from the SNOs I see here that they’d be very happy to add a new disks as old ones fill. Bring it on! :slight_smile:

2 Likes

Do I see correctly that recent test data is stored with TTL of 14 days?

1 Like

I’ve only ever seen 30 days mentioned. Do you have a link?

I just checked a node where I deleted databases 4 days ago. Almost all TTL data is 30 days.

1 Like

I’m looking at my nodes’ databases.

select piece_expiration from piece_expirations order by rowid desc limit 10;

2024-07-07 01:44:56.581389+00:00
2024-07-14 01:44:34.74594+00:00
2024-07-14 01:44:48.675778+00:00
2024-07-14 01:44:45.278975+00:00
2024-07-14 01:44:41.918976+00:00
2024-07-14 01:44:50.715372+00:00
2024-07-14 01:45:10.172953+00:00
2024-07-14 01:44:34.042397+00:00
2024-07-14 01:45:10.252893+00:00
2024-07-14 01:44:59.720131+00:00
1 Like

Is that maybe test data uploaded 2 weeks ago?

order by rowid desc shows the most recently added entries.

About two weeks ago littleskunk mentioned some TTL data accidentally got sent to the trash (reusing filenames is like a delete+reupload I guess?). Perhaps they’re filling that hole so they get the amount of TTL-expiring data each day as they want?

1 Like

Is it from a Saltlake satellite’s customers?

exactly, unfortunately…
Since it’s being deleted, well, it will use the trash folder.

Yes that observation is correct. No deals are signed yet but we have some possible start dates. The challenge is to get enough throughput on day 1. Adding surge nodes takes a bit of time. Reducing the amount of testdata we are allocating will buy us time. It is not a solution. The customers are going to catch up and we end up in the same situation as we are now. But we can use that extra time to order more surge nodes.

There was also a debate internally why we don’t stop uploading. That would be even better for surge node planning. But there is a trust issue. We don’t want you to lose trust. So for now we have to continue uploading. It might change the moment deals are getting signed.

9 Likes

Nothing to lose. Stop those fake data uploads.

2 Likes

Just untrust SLC and the uploads are going to stop for you.

3 Likes

Did somebody say more SLC data for me? Because I’m OK with more SLC data for me. Totally down with it. I’m chill. We cool. Just keep sending those ones and zeroes…

1 Like

also related payout to your wallet…
Even yours, @pangolin
But well. It’s fully upper on you - make it untrusted and the problem is solved.

1 Like

If customers don’t sign this will stop anyway. SNOs knowing that with or without fake uploads.
.

2 Likes

You know there is a signature on the paper from both sides. Just think about the conditions that are required for the company to sign these deals. Stopping the uploads is the equivalent of telling no these customers even if they would like to use us.

2 Likes

Out of curiosity—do the prospective customers follow our discussions here? Are they happy for honesty, or worried about SNO behavior?

They’re probably looking to see if they’re getting the performance they expect for the money they pay: and don’t care about a few yahoos in a forum. People may talk about a lot of stuff: but just watch what their actions are when they’re getting paid :wink:

6 Likes

Although there are currently 11000 full nodes, my ingress is the same as when there where 8000 full nodes. I still don’t get why ingress didn’t increase even a little bit.
I spread the load over multiple nodes, each node have its own disk, plenty of memory and cpu resources. And to reduce the iops, I moved the databases and log file to an other disk. But ingress is still hovering around 110mbit.
Also checked the successrate. Overall I have a (ingress) successrate of 97.x%.

Am I just unlucky that I’m to far away? Are others closer by and win the races more often? Or is it just that there is not enough data uploaded onto the network to fully utilize my bandwidth?

Overall for this month I got 25,5TB of ingress (measured on my storj vlan in pfsense), I’m not entire sure how much extra disk space is still occupied compared to last month. But not to shabby for the current load, but I wouldn’t mind if it would be much more…
Oh, and I don’t mind that it is test data.

2 Likes

Maybe my expectations are just wrong. But when there are concerns about customers upload speeds, I though that my ingress load would be at least a bit higher