user registration while a wallet is disconnected (but is now up) is not handled well

[decred]

user registration while a wallet is disconnected (but is now up) is not handled well

I have to restart dcrstakepool 3-5 times per day.
On this hype number of registrations reached 1000 users and it goes down very often after new registration.

Currently, I'm making a bash script which will check page status and restart the service if it's not 200. But I hope we can find a better solution than restart and resync it so many times per day.

[jolan]

OK I'll give it another look soon (once the release work for 0.8.0 is done with). I did look at it a couple weeks ago but had difficulty reproducing it on my unloaded testnet/dev pool which made it hard to debug.

[chappjc]

I'm running without the smarter connectivity PR changes, but when I tried it on testnet ages ago, I think the status URL triggered reconnect

[chappjc]

From an admin IP that is

Actually, it seems that this time it happened after failed stakepooluserinfo request.

What I see in the dcrstakepool.log:

18:20:29 2017-02-08 [TRC] DCRS: Sending command [stakepooluserinfo] with id 6
18:20:31 2017-02-08 [ERR] DCRS: Websocket receive error from XXX:9110: unexpected EOF
18:20:31 2017-02-08 [TRC] DCRS: Disconnecting RPC client XXX:9110
18:20:31 2017-02-08 [TRC] DCRS: Shutting down RPC client XXX:9110
18:20:31 2017-02-08 [TRC] DCRS: RPC client input handler done for XXX:9110
18:20:31 2017-02-08 [TRC] DCRS: Sending command [stakepooluserinfo] with id 77
18:20:31 2017-02-08 [TRC] DCRS: RPC client output handler done for XXX:9110

And wallet.log:

18:20:32 2017-02-08 [WRN] RPCS: Websocket receive failed from client 81.2.253.91:45685: read tcp4 XXX:9110->81.2.253.91:45685: use of closed network connection

18:20:32 2017-02-08 [WRN] RPCS: Failed websocket send to client 81.2.253.91:45685: write tcp4 XXX:9110->81.2.253.91:45685: i/o timeout

18:20:32 2017-02-08 [INF] RPCS: Disconnected websocket client 81.2.253.91:45685

Wallet was online all this time and continue working ok. I see some voted tickets right after this sockets error. Of course, stakepool will never try to reconnect, so now only 2 wallets are still connected.

[chappjc]

Pretty sure going to /status from an admin IP is supposed to reconnect

[jholdstock]

@dajohi I believe this can be closed thanks to #494

[JoeGruffins]

@jholdstock ok to close this?