Awesome running 6 vaults around the office and at home
Was there a bump downwards in the number of nodes in the table just now, or was it just my cloud node? Indirect connections suddenly dropped from 25 to 3, while the table size went from 78 to 63.
Do we have performance issues at the moment? Safe launcher is still slow, and demo app now registered for access rights, launcher granted them. all right. then demo tried to load the initial stuff ever so slow and staying greyish and eventually the launcher went back grey and displays (but greyed/blocked) the access rights stuff again, weird never seen it this way before)
Also before when I was with the same user on the network the demo app went further through to upload a www but always failed with error to create /public/www no matter if i went template or local files.
Also i noticed, the launcher often initially when starting up tcp syn bursts to lot of remote ip locations, which briefly brought down the local net here (sucky consumer grade cpe at various places, this might become a serious showstopper in mass deployment with crappy grade consumer broadband providers and their cpe (customers premises equipment)
Anyways when it finally settled, I was running a vault on a different machine on the lan and I have seen the launcher only connect (longterm) to this vault ip on the lan and nowhere else. Could this be some of the reasons why the demo then has problems of getting packets and meta and chunks to and from the net (going only through this local vault?)
Thanks.
P.S. Hmmm, and now the demo even says could not connect to launcher when trying to load its initial stuff. Okay and now the lan or for this launcher/demo machine is saturated (tcp syn wise). Could we please have a more relaxed attempt in low level ip/tcp stuff. Thanks.
Circa 25% of the network dropped, so somebody running a lot of vaults (net is over 850 vaults for a while) dropped them all. Lets see what that does in a smallish network now.
Anyone able to acces .safenet links ?
My vault is running fine,printing lots of logs, but I canât access my page, consisting of a 0.3kb index.htmlâŚ
No, seems like it was closer to 40 - 50% dropped so several hundred nodes at once. Looking at log now to see what I can work out. Under 3000 nodes the tree is not well balanced so interesting to see this.
We have cache and bootstrap cache disabled here.
As soon as I close the launcher, the local machine (win7, x64) can e.g. ping destinations again. Running the launcher for a while here apparently drives the ip stack of windows into shutdown state
@dirvine how are you able to determine the number of vaults, just wondering?
We have not seen that one, perhaps parallel bootstrap, be good to get your log file there.
We can run a script to work out number of unique idâs seen on all the droplets we run and add them up. Itâs very unscientific, but close for small networks though So if we see 850 nodes there are likely several more, as network grows so will that error rate though.
Was running 8 vaults earlier while using VPN. Registered on the network and uploaded the demo page. Then restarted and now running one vault without VPN. Logging in to launcher easily with previously created credentials. When starting the demo app it asks for permission in the launcher. I click allow and the launcher just shows the busy circle and the maidsafe demo app just gives up after a while giving error message can not connect to launcher, should keep launcher running. Closing both applications, restarting and retrying returns the same result over and over.
I see an interesting difference between the logs of my cloud vault and local vault: The local vault has a line that says âstate changed to clientâ while the cloud vault has nothing like that. Is that a new adaptation to bandwidth on the respective machines?
Also, the local vault (state=client) reports âclient connectionsâ in its log, while the cloud vault has no entries like that. Does that represent a division of labor between high-bandwidth vaults and lower-bandwidth vaults?
This would be good, of course, and something discussed previously (adaptation to bandwidth conditions).
I noticed this when i wondered why the cloud vault had no client connections, which it has always had previously.
Crust also allows for blacklist support during bootstrap so that Routing can specify which peers are not advisable to connect to because for example it concluded that this peer had suboptimal Routing table size.
Curious: Whatâs the scenario here? Is the routing table dynamic based on put/gets⌠is this about a simple task vs a complex task in the network affects the size of a peerâs routing table? I donât understand how and why each peerâs routing table changes size.
Why so long? I thought things were improving?
That is the binary tree balancing really. As nodes join they affect groups around them and with exponential decrease affect nodes far from them. Many nodes wont see them at all, but many will affect their routing table to balance as this new node joins or leaves.
Anyhoo, I am for a nap so will be back in a few hours, been a frantic week or two here. Really keen to see the routing network here and allow us to focus on data now we look like having resources cracked (in terms of cpu/bw I mean) Lets see though.
Hi All,
My site took a while to upload and feels slower when I navigate to it than it did in Testnet3. Does anyone else have a page I can check the speed on?
cheers all
Al
Where are files stored locally on Windows System ?
my site doesnât want to load now, but I am also uploading small pic files to my drive at the same timeâŚ
I suspect we still have a launcher bug as well here. If so it will get fixed quickly This is what the tests are for really. I need to rifle the logs though, to know for sure. It could be the massive drop in vaults though, there is not much resilience in these tests for that, so double edged sword, make running lots of vaults easy, but we need to put in the resilience as well in case folk run hundreds in small network tests. So yin and yang perhaps.