Multichain does not shutdown properly on kill <pid> anymore

Question

Hi I experience the problem that multichain does not shutdown properly anymore.

 

When I do a kill <mc-pid> the chain is not shutdown. When I look into the logs it says:

2017-10-06 11:21:06 msghand thread interrupt
2017-10-06 11:21:06 net thread interrupt
2017-10-06 11:21:06 Shutdown: In progress...

What I can remember to have changed: I activated rpcssl feature.

But don't know it this has something todo with the problem.

 

Alex

Comments

Thanks - I'll forward this to the team and we'll take a look.

Answer 1

We've looked into this and it does not appear to be caused by the rpcssl option, because we could not reproduce it with or without that option. There are some other causes that we are now looking into. Could you please let us know if the node that didn't shut down properly was one of the miners of the chain?

Comments

Hi.
Setup was single node.

---

<EDIT>: Sorry, didn't know I by accident posted it here.
See: https://www.multichain.com/qa/7006/node-rpc-completely-freezes-rpcssl

There is another (maybe related) severe problem. It is not only impossible to stop the node via kill <pid>, the node also does not react anymore on any rpc call. It runs into a timeout, even if I do a local "curl". The last debug entries were:

2017-10-09 09:03:20 MultiChainMiner: Block Found - 0007102ae08c859c49970a9951a0660637ba6bc3c465d09fe384aaac7a763410, prev: 00ddb3a5ac273f3525a73ebf5f4f2fb19f327f4a1e6229ee37d038dd5b648700, height: 126, txs: 1
2017-10-09 09:03:20 UpdateTip:            new best=0007102ae08c859c49970a9951a0660637ba6bc3c465d09fe384aaac7a763410  height=126  log2_work=14.988685  tx=140  date=2017-10-09 09:03:20 progress=1.000000  cache=0
2017-10-09 09:03:27 ResendWalletTransactions()
2017-10-09 09:03:51 MultiChainMiner: Block Found - 001296bcc218cd470008d2c2e88372c586f8d756bcc06962ad7c5483b8cf290d, prev: 0007102ae08c859c49970a9951a0660637ba6bc3c465d09fe384aaac7a763410, height: 127, txs: 1
2017-10-09 09:03:51 UpdateTip:            new best=001296bcc218cd470008d2c2e88372c586f8d756bcc06962ad7c5483b8cf290d  height=127  log2_work=15  tx=141  date=2017-10-09 09:03:51 progress=1.000000  cache=0

And after a kill <pid> the debug.log stucks here:
2017-10-09 10:58:44 dumpaddr thread stop
2017-10-09 10:58:44 msghand thread interrupt
2017-10-09 10:58:44 addcon thread interrupt
2017-10-09 10:58:44 opencon thread interrupt
2017-10-09 10:58:44 net thread interrupt
2017-10-09 10:58:44 Shutdown: In progress...

---

Noted, and we'll look into it.

FYI if you *need* to stop the node on the operating system level, you can use "kill -9" rather than just "kill".

Second, for now is there a reason not to just use the 'stop' API command?

---

I was just wondering why kill without -9 does not work anymore.
I will try the stop command, if this shuts down properly.

---

Yes, seems to work. I will use it instead of kill.
Thank you

---

Great. You should definitely use the stop command as the preferred approach. But we're also going to take a detailed look at this problem so that 'kill' is supported as well.