Hello, I unfortunately forgot the saltlake Satellite by mistake. Is there a way to get the Satellite back? I ran this command thinking, it will forget the right satellites, but in the end I messed up and “forgot” the wrong one. Please help. Thanks in Advance.
You can check its progress using forget-satellite-status command.
I don’t know if you got my problem. I unfortunately forgot/removed the saltlake satellite, and want it back.
I understood your problem both here and other thread. The status command might help you figure out if the command executed and finished or still underway. I am pretty sure, once executed it can’t be reverted.
1 Like
Yes, it executed, and it doesn’t took long (around 10-15 minutes) to get a mail regarding suspension from saltlake satellite, because none of the audits were successful, so I shut down the node until I can fix it.
Unfortunately you cannot return it back (Russians have a saying: “it is impossible to turn the minced meat back”). Even if it would be possible to do manually, all related data is already deleted, right?, so it will be immediately disqualified.
You may only start over, or - start another node on a separate disk with a new identity signed with a new authorization token.
Depends on if the Node is already “minced” in my case. I asked Copilot in the meantime, and it told me to add the Satellite manually to the config.yaml. I found a tag with “trusted satellites” with a link in the yaml. When opening it it seem to lead to a storj site where all 4 satellites are mentioned with ID and Name, nothing more. I uncommented the line with the trusted satellites, started the node, and the Saltlake satellite one came back to life. My Suspension dropped in the 10-15 min timespan to around 78% yesterday, as I “forgot” the satellite. Now it grew back to 100%, I also got a mail, that the node is now “unsuspended” on saltlake. So it seem to run perfectly again. It’s great what copilot and yaml inspecting can do.
This will not help, if the command is confirmed. You may also do not change anything in your config.yaml at all, it will load the default trusted satellites list.
However, if you did a “forget-satellite” command and it’s succeed, there is no way back, I’m sorry.
In my case it seemed to be in “processing” state. I noticed it relatively quick, so I stopped the node, and “fixed” it, by unchecking the line in the yaml file. I hope this will survive the upcoming node updates. In my case the Saltlake satellite is now back in the dashboard showing 100% suspension rate, which was missing after I “forgot” the satellite. I also deleted the container and re-pulled everything. So it is working as usual at the moment.
Did you lose any data since the command has been issued?
No, when I was executing the command, I could see errors with “satellite is not trusted” in the docker locks for “downloads” which is uploading from my end, with the saltlake ID. After a few minutes, I got a disqualification note, and saw that saltlake was missing in the node dashboard. To prevent getting disqualified, I shut down the node, and looked for a possible fix. I did the yaml thing, and restarted my node. Saltlake was back in the dashboard again, and I got a mail again, that it is now unsuspended. Everything seem to work fine since then. The “forgotten” status was just for a few minutes, and not hours or days. Looks like I did not lost any data. Maybe I was lucky and reverted it fast enough, before it could do further damage. Time will tell.
Do you mean suspension ?
Yes, I meant that, sorry, my mistake.
And I think there is some sort of lesson for us here… 
If the forum gets filled more with usefull information and fixes, copilot could help most issues in short time. I like the forum, but some threads have over 1k comments. I want to fix an issue and not reading a tale 
Hahahaha! That made me chuckle. Point taken
Nah. I gave up on LLMs beyond trivial common knowledge scenarios.
Even after a lot of prompt engineering and walking on eggshells, even gpt4 and gemini produce beautiful and eloquent, quite convincing and sometimes even plausible horseshit, for anything that is even remotely non-trivial.
They this whole step by step explanation, with step 5 being complete bullshit. Then you point I out that that’s bullshit. They go “of course!”, politely explain back to you why it is bullshit and produce a replacement horseshit. Repeat until their entire essay turns to horseshit or it stumbles out of your competency so you can’t verify.
It’s a massive waste of time. Machine learning is great in niche well defined applications, but to get to an all-knowing general AI as it is (mis) represented — we need another 300 years of progress
The better approach is to ask it to summarize existing text. If there was an easy way to copy text from a mile long discourse thread it would have been even better. But then you still have to verify.
3 Likes
THIS is what scares me the most, if I’m honest.
It’s only I ask them about things in my field of expertise that I realise how much tripe can come out that sounds very convincing.
2 Likes
Your node could be lucky then, or you didn’t provide a --force flag (it will eliminate all related data from the data location too, not only from databases).