bitcoind not accepting connections on 8333

[krile]

Hello all,

I decided to run a full node to help out the network. I have a fast fiber connection, so I guess it would make a good contribution.

Everything already worked, but then I had to restart my server after a few days and since then, bitcoind does not accept connections anymore - it is still connected to the network and is updating the new blocks as they come...

Here is some more info:

I am running version 0.10.0 compiled from source on ubuntu server 14.04, no GUI, so only CLI tips please.

bitcoin.conf

Code:

server=1
daemon=1
rpcuser=bitcoinrpc
rpcpassword=xxxxxxxxx
rpcport=8332

bitcoind debug.log

Code:

2015-03-31 20:48:05 Bound to [::]:8333
2015-03-31 20:48:05 Bound to 0.0.0.0:8333
2015-03-31 20:48:05 init message: Loading block index...

netstat --ip -lpa gives me:

Code:

tcp        0      0 localhost:8332          *:*                     LISTEN      1533/bitcoind   
tcp        0      0 *:8333                  *:*                     LISTEN      1533/bitcoind   

A port scanning website gives me this:

Code:

My IP x.x.x.x is responding on port 8333 ().

I use a direct PPPoE connection, no router in between. And from the info provided above, my IP is reachable on 8333.

However according to blockchain.info and also https://getaddr.bitnodes.io/ my node is not reachable.

I googled all the options, but mostly people have problems with routers and port forwarding etc. Here the port is reachable, so it has to stop somewhere in bitcoind. The bitcoin.conf file does not have any options to even try solving this...

And btw, its not a general issue on the server, I can connect via SSH (21) from the outside normally and also from other ports...

Any help would be appreciated.

[2112]

I use a direct PPPoE connection, no router in between. And from the info provided above, my IP is reachable on 8333.

You maybe suffering from an MTU issue. PPPoE has MTU 8 bytes less than the Ethernet standard of 1500 bytes.

See if you can reconfigure your modem from PPPoE to PPPoA which would then support full 1500 byte MTU.

I don't have time to give you a full lecture on the MTU, path MTU discovery, MSS, etc.

Start here: http://www.dslreports.com/tweaks

Edit: If your ISP gave you a NAT router/gateway and you reconfigured it to the bridge mode, then most likely this router was transparently adjusting MSS to be 8 bytes lower to accommodate PPPoE overhead.

[krile]

I use a direct PPPoE connection, no router in between. And from the info provided above, my IP is reachable on 8333.

You maybe suffering from an MTU issue. PPPoE has MTU 8 bytes less than the Ethernet standard of 1500 bytes.

See if you can reconfigure your modem from PPPoE to PPPoA which would then support full 1500 byte MTU.

I don't have time to give you a full lecture on the MTU, path MTU discovery, MSS, etc.

Start here: http://www.dslreports.com/tweaks

Edit: If your ISP gave you a NAT router/gateway and you reconfigured it to the bridge mode, then most likely this router was transparently adjusting MSS to be 8 bytes lower to accommodate PPPoE overhead.

Thank you for the advice. If the MTU issue is the case, how does that explain the fact that the node was reachable normally for days. The problem occurred after a planed server restart.

I tried some different MTU settings but still not reachable, but the port is reachable from the outside...

Also tried removing bitcoind and installing it from the ppa:bitcoin/bitcoin. Still does not work.

Any other advice anyone?

[btchris]

Thank you for the advice. If the MTU issue is the case, how does that explain the fact that the node was reachable normally for days. The problem occurred after a planed server restart.

I tried some different MTU settings but still not reachable, but the port is reachable from the outside...

Also tried removing bitcoind and installing it from the ppa:bitcoin/bitcoin. Still does not work.

Any other advice anyone?

Are you OK posting (or if not posting, at least PMing) your public IP?

[krile]

Thank you for the advice. If the MTU issue is the case, how does that explain the fact that the node was reachable normally for days. The problem occurred after a planed server restart.

I tried some different MTU settings but still not reachable, but the port is reachable from the outside...

Also tried removing bitcoind and installing it from the ppa:bitcoin/bitcoin. Still does not work.

Any other advice anyone?

Are you OK posting (or if not posting, at least PMing) your public IP?

PM sent

[btchris]

Are you OK posting (or if not posting, at least PMing) your public IP?

Got your PM. It's not obvious what's going on, though.

Confirmed that your node is accepting connections on TCP 8333, 3-way handshake succeeded.

After the initial connection, my node sent a version message as expected. Your node ACKed receiving the message at the TCP level.

At this point, your node should have sent it's own version message to mine, and it should have ACKed the version message at the app level with a verack message. My node received neither.

There are no MTU issues (yet) -- all of the initial messages are far under 1481 bytes (the point at which MTU issues might rear their ugly head on an older PPPoE connection).

There was no obvious evidence of a transparent proxy (no differing TTLs), but it's always possible.

What Bitcoin node software are you running? Do you have any familiarity with tcpdump (I'd be interested if your node was responding at all, and if the response was getting lost, or if it's not responding at all)?

Anything interesting in the logs?

[krile]

Are you OK posting (or if not posting, at least PMing) your public IP?

Got your PM. It's not obvious what's going on, though.

Confirmed that your node is accepting connections on TCP 8333, 3-way handshake succeeded.

After the initial connection, my node sent a version message as expected. Your node ACKed receiving the message at the TCP level.

At this point, your node should have sent it's own version message to mine, and it should have ACKed the version message at the app level with a verack message. My node received neither.

There are no MTU issues (yet) -- all of the initial messages are far under 1481 bytes (the point at which MTU issues might rear their ugly head on an older PPPoE connection).

There was no obvious evidence of a transparent proxy (no different TTLs), but it's always possible.

What Bitcoin node software are you running? Do you have any familiarity with tcpdump (I'd be interested if your node was responding at all, and if the response was getting lost, or if it's not responding at all)?

Anything interesting in the logs?

Its running the latest version of bitcoind, that is 0.10.0.

In the logs I am only getting block updates.

Like mentioned above, on startup of bitcoind I get:

Code:

2015-04-01 20:34:21 Bound to [::]:8333
2015-04-01 20:34:21 Bound to 0.0.0.0:8333
2015-04-01 20:34:21 init message: Loading block index..

Then it just updates the blocks normally...

[krile]

Well we will never know what was wrong. I wiped the server and reinstalled everything.

The .bitcoin folder is the only thing that I migrated.

It works now as it should. But now I'm afraid to restart the server though

Thank you for all the efforts.

[vr4dude316]

hello.  i am having the exact same issue.  all port checking tools reveal 8333 is open to the public.  bitnode21 checks ok that server is reachable, but status shows as down.
what gives?

[achow101]

hello.  i am having the exact same issue.  all port checking tools reveal 8333 is open to the public.  bitnode21 checks ok that server is reachable, but status shows as down.
what gives?

Are you fully synced?