WOW! I don’t believe it because I personally noticed dozens of nodes without free space, which the satellite gave me. HOWEVER, I have to admit that I was wrong (since my guesses are subjective, and your data is objective) and the number of nodes exceeded my expectations at times.
Please note, that our selection code is not perfect yet. By any means!
A simple example as it currently works still:
Your node is advertising the same space to all satellites across the network, which uses that value until the next report comes in. In that interval, all satellites can hammer your node full with uploads and the given satellite would still select your full node as destination for repair puts/ graceful exit transfers and normal uploads.
We are slowly making progress. Rome was also not built in one day
For what it’s worth, each piece transfer during graceful exit is tried 5 times on different nodes. So the chances of failing a piece even if you fail 35% of transfers is about 0.5%. Since you can fail up to (I believe) 10% on graceful exit, this is not going to effect the end result. Though the last in depth report didn’t show anywhere near the percentages you’re mentioning. As far as I know it’s closer to 10% transfer failures. At that point there is a 0.001% chance of a piece transfer failing all 5 tries.
sure burned in a day tho… xD
but i suppose that’s why they invented the fire-brigade…
also it’s easy to have 8k nodes when some people are on 8-10 different ip/subnets just so they can exploit the current and very poorly designed ip/subnet nodes data allocation scheme… a guy like that might be running 40-80 nodes on one or two servers…
There is though, since each upload starts 110 uploads there is plenty of overhead to deal with a few failed transfers. As Stefan said, it could still be improved, but shouldn’t be much of a problem already.
If you reperpes your sattelite do you change it’s ID and make it clean, as lot of peple made GE from there, as was told it will be shuted down, and there was lot of money in held amount than can be used for upgrade.
I can say for sure, that my satellite ID will not change.
That said, i or more specifically we at Storj are seeing what the best options are to “reset” it.
First of the tasks is to pay out the held amount, then the rest can be discussed/done.
Are there approximate “reset” date? Of course I done GE and now just wondering what will happen with held amount. Will the rest be paid 50%?
And if it’s not difficult, you can clarify my yesterday’s question about the number of nodes and give a little more information: not only the number of nodes, but also the number of unique “groups of nodes” in subnets /24? I hope I’m not asking too much
I can not tell you an date yet. The withheld money, that we owe active nodes on my satellites is paid out as a “final sum”. Basically no outstanding payments after that date.
In terms of clarification around nodes, I need to check which information I am allowed to hand out.
That said, i am pretty certain we gave some metrics in the last town hall.
Some have even cautioned against doing so all along. I know I have. It’s always been a bit of a gamble to GE a still active satellite. I can imagine the held back amount on this satellite looking mighty tempting, but especially after 50% being paid out already now, I’ll gladly wait to see what’s next for it instead. And that has always been my position at least.
I know I didn’t
On an IT perspective, it never made much sense to me to shut it down…so, it’s active on my nodes alright, and a lot of $$$ in held amount to
It is a legit gamble if you run one node per HDD because for older nodes the held amount was over 100-200$ depending on the node size, meaning you could buy another HDD from that money. But if your current HDD fails and you lose that node, you get nothing.
That’s why I did it. Now that I have a raidz1 I wouldn’t do it again but that’s too late now. At the time my node was running on an old HDD that could have died anytime.