rpc tests: increase http timeout #8052

pull kazcw wants to merge 1 commits into bitcoin:master from kazcw:rpcwallet-test-timeout changing 1 files +1 −1

kazcw commented at 5:15 PM on May 13, 2016: contributor

30s is not enough. An importwallet test case can take up to about 40s my test machine.

Fixes #8051.

rpc tests: increase http timeout

30s is not enough. An importwallet test case can take up to about 40s my test
machine.

Fixes #8051.

53bcca971b

jonasschnelli added the label Tests on May 13, 2016
jonasschnelli commented at 5:18 PM on May 13, 2016: contributor

Ah. This makes sense! Fixes #8045 not #8051. I could not reproduce #8045 but looks good.

utACK 53bcca971b807b25f219f0aeea25071e38ccea7d
sipa commented at 7:11 PM on May 13, 2016: member

This does not fix #8045 for me.
kazcw commented at 7:41 PM on May 13, 2016: contributor

Hm. I can't reliably reproduce the test failure; I got it to happen once by looping walletbackup.py until it failed, but the problem's too intermittent here for me to test a solution effectively.

In the error case, packet capture showed: t+0s: importwallet request to server t+30.5s: FIN+ACK from client t+38.5s: reply from server, followed by FIN+ACK from server

So it looks like it involves a 30 second timeout somewhere on the client end, but since 53bcca9 didn't fix it apparently the problem timeout's value is not originating with HTTP_TIMEOUT...

[[I should probably modify this to extend the timeout only for the test case in question, but let's figure out why it isn't working first]]
sipa commented at 7:54 PM on May 13, 2016: member

I tried with HTTP_TIMEOUT=600, and walletbackup.py still fails reliably. #8051 fixes it.
sdaftuar commented at 7:57 PM on May 13, 2016: member

Hmm, I think this might be related: #6695 @sipa Does it fix it for you if you start bitcoind with a bigger -rpcservertimeout?
kazcw commented at 8:03 PM on May 13, 2016: contributor

Oh, the bug in #8051 occurs during send. This fixes a bug in receiving...
sipa commented at 9:36 PM on May 13, 2016: member

@sdaftuar That works!
MarcoFalke commented at 6:31 AM on May 15, 2016: member

I tried with HTTP_TIMEOUT=1, and walletbackup.py still works fine.

kazcw commented at 3:17 PM on May 15, 2016: contributor

This is the error (after INFO: Restoring using dumped wallet):

Unexpected exception caught during testing: timeout('timed out',)
  File "/bitcoin/qa/rpc-tests/test_framework/test_framework.py", line 138, in main
    self.run_test()
  File "./walletbackup.py", line 188, in run_test
    self.nodes[0].importwallet(tmpdir + "/node0/wallet.dump")
  File "/bitcoin/qa/rpc-tests/test_framework/coverage.py", line 49, in __call__
    return_val = self.auth_service_proxy_instance.__call__(*args, **kwargs)
  File "/bitcoin/qa/rpc-tests/test_framework/authproxy.py", line 137, in __call__
    response = self._request('POST', self.__url.path, postdata)
  File "/bitcoin/qa/rpc-tests/test_framework/authproxy.py", line 119, in _request
    return self._get_response()
  File "/bitcoin/qa/rpc-tests/test_framework/authproxy.py", line 152, in _get_response
    http_response = self.__conn.getresponse()
  File "/usr/lib64/python3.4/http/client.py", line 1171, in getresponse
    response.begin()
  File "/usr/lib64/python3.4/http/client.py", line 351, in begin
    version, status, reason = self._read_status()
  File "/usr/lib64/python3.4/http/client.py", line 313, in _read_status
    line = str(self.fp.readline(_MAXLINE + 1), "iso-8859-1")
  File "/usr/lib64/python3.4/socket.py", line 374, in readinto
    return self._sock.recv_into(b)
Stopping nodes
WARN: Unable to stop node: CannotSendRequest('Request-sent',)

It doesn't happen often when my machine is otherwise idle; I looped the test for several hours (originally trying to reproduce the other bug) and got two failures. With the machine under heavy load it happens consistently (all failures look just like the above backtrace).

When I set HTTP_TIMEOUT=1 it fails in an earlier recv_into (in various rpc calls).

laanwj commented at 1:49 PM on May 31, 2016: member

As this doesn't fix the issue, we should probably close it?

If we need this I'd prefer to not change the default value in authproxy.py but supply a different timeout where the object is created instead.
kazcw commented at 5:40 PM on June 1, 2016: contributor

Yeah, this is too intermittent to worry about; it only came up because I was grinding the test to look for that other bug.
kazcw closed this on Jun 1, 2016
MarcoFalke locked this on Sep 8, 2021

Contributors

Labels

Linked (view graph)

#8051 Seemingly fix walletbackup.py failure