BEWARE forked or half dead altcoins that will eat up your download limit

[almightyruler]

A week ago something on my home office network started using a heap of bandwidth. My average daily download went from 4GB/day to 28GB/day. By the time I noticed a few days later my monthly download was about 120GB higher than it usually is at this point of the billing cycle.

What happened? There is a bug/oversight in *coin code that causes a client to misbehave when a fork occurs; the client repeatedly asks for the same blocks from peers on another fork, but of course rejects them because they don't match the local copy of the chain. Client A doesn't slow down its rate of requests to Client B, even though it is getting back exactly the same data each time and will never resolve the fork, plus Client B does not ban Client A for abuse either. It's just two stupid computers stuck in an endless loop.

I think in this case it was either grain or mintcoin that caused the extra 120GB of download, but it can happen to pretty much any coin when it forks. I'm running over 60 altcoin clients so I guess I need to keep a better eye on things.

tl;dr if you have a download limit and leave old clients running you may end up in some trouble.

[raskul]

is the coin you were mining not finding blocks?

[almightyruler]

is the coin you were mining not finding blocks?

Nothing to do with mining, this is just the client. The *coin code does not seem to be very good at resolving a fork situation.

[raskul]

is the coin you were mining not finding blocks?

Nothing to do with mining, this is just the client. The *coin code does not seem to be very good at resolving a fork situation.

ah sorry, misread. gotcha

[fernando]

is the coin you were mining not finding blocks?

Nothing to do with mining, this is just the client. The *coin code does not seem to be very good at resolving a fork situation.

Just curious... why do you keep then open?

[brioche]

is the coin you were mining not finding blocks?

Nothing to do with mining, this is just the client. The *coin code does not seem to be very good at resolving a fork situation.

Just curious... why do you keep then open?

supporting the network likely

[almightyruler]

is the coin you were mining not finding blocks?

Nothing to do with mining, this is just the client. The *coin code does not seem to be very good at resolving a fork situation.

Just curious... why do you keep then open?

The clients are running on a dedicated *nix server, not a Windows box. Coin clients mostly consume very little bandwidth or CPU (so long as PoS is disabled), and most of the RAM they use is inactive so it's swapped out. I think I could comfortably run 150 to 200 daemons on this box.

Keeping the clients active helps the network of each of the coins. The smaller coins would quickly die if there were no nodes that were available most of the time.

[sgtstedanko]

You didn't have Dafuq coin installed by any chance?

[Amph]

You didn't have Dafuq coin installed by any chance?

that shit tell you to run the client as admin right?

[almightyruler]

You didn't have Dafuq coin installed by any chance?

that shit tell you to run the client as admin right?

Nope, never heard of that coin. I generally only run clients that I have compiled myself.

[platorin]

I'd stay away from anything that looks suspicious.

[hvezdasmrti]

Good... There are countries with limited bandwith ahahahaha, 19th century.

[moffer5]

I'd run an antivirus scan if I were you.

[almightyruler]

I'd run an antivirus scan if I were you.

I explained in the OP that it's a bug which causes the client to get stuck in an endless loop, repeatedly requesting the same set of blocks. I've seen this happen before, although it's usually the other client which is going bonkers and repeatedly asking mine for blocks.

It may have been fixed in some newer maintained coins, but the buggy code is still sitting there in the few hundred other coins...

The coin client hogging my bandwidth was most likely grain, because right now it has a very low difficulty, which increases the chances of forking.

[binaryclock]

Shitty coin code?

NO!

I don't believe it.

[Chicago]

Hello,

    I too have observed this behavior with 0.8 branch clones of the Satoshi reference client.
    Its most prominent when a new wallet starts syncing from block zero and approaches the tip.
    It even happens when a bootstrap.dat is used to get 90% of the way there.

    While observing the behavior, I note the client *eventually* becomes convinced it is on the right chain and stops requesting blocks over and over across the network.
    I've also noted that after stopping the client and then starting it again, the behavior does not repeat immediately; however, if you leave the client off for a week and then start to sync it will happen.

    Since I have been working with a few coin revivals on the 0.8 branch, I'm looking forward to the first one rebasing against Bitcoin Core's 0.10 branch where hopefully the behavior disappears when the blockchain must be downloaded and built anew.

    Also, although I don't have a link to share - while researching this behavior earlier in the year, I stumbled across an IRC discussion with one of the Bitcoin Core developers noting the same oddity on Bitcoin Mainnet.
    This is reassuring in that it convinces me life can go on despite the quirky recursion through forks.

Best Regards,
-Chicago