History

Wladimir J. van der Laan a88bd3186d Merge #14670 : http: Fix HTTP server shutdown `28479f926f` qa: Test bitcond shutdown (João Barbosa) `8d3f46ec39` http: Remove timeout to exit event loop (João Barbosa) `e98a9eede2` http: Remove unnecessary event_base_loopexit call (João Barbosa) `6b13580f4e` http: Unlisten sockets after all workers quit (João Barbosa) `18e9685816` http: Send "Connection: close" header if shutdown is requested (João Barbosa) `02e1e4eff6` rpc: Add wait argument to stop (João Barbosa) Pull request description: Fixes #11777. Reverts #11006. Replaces #13501. With this change the HTTP server will exit gracefully, meaning that all requests will finish processing and sending the response, even if this means to wait more than 2 seconds (current time allowed to exit the event loop). Another small change is that connections are accepted even when the server is stopping, but HTTP requests are rejected. This can be improved later, especially if chunked replies are implemented. Briefly, before this PR, this is the order or events when a request arrives (RPC `stop`): 1. `bufferevent_disable(..., EV_READ)` 2. `StartShutdown()` 3. `evhttp_del_accept_socket(...)` 4. `ThreadHTTP` terminates (event loop exits) because there are no active or pending events thanks to 1. and 3. 5. client doesn't get the response thanks to 4. This can be verified by applying ```diff // Event loop will exit after current HTTP requests have been handled, so // this reply will get back to the client. StartShutdown(); + MilliSleep(2000); return "Bitcoin server stopping"; } ``` and checking the log output: ``` Received a POST request for / from 127.0.0.1:62443 ThreadRPCServer method=stop user=__cookie__ Interrupting HTTP server ** Exited http event loop Interrupting HTTP RPC server Interrupting RPC tor: Thread interrupt Shutdown: In progress... torcontrol thread exit Stopping HTTP RPC server addcon thread exit opencon thread exit Unregistering HTTP handler for / (exactmatch 1) Unregistering HTTP handler for /wallet/ (exactmatch 0) Stopping RPC RPC stopped. Stopping HTTP server Waiting for HTTP worker threads to exit msghand thread exit net thread exit ... sleep 2 seconds ... Waiting for HTTP event thread to exit Stopped HTTP server ``` For this reason point 3. is moved right after all HTTP workers quit. In that moment HTTP replies are queued in the event loop which keeps spinning util all connections are closed. In order to trigger the server side close with keep alive connections (implicit in HTTP/1.1) the header `Connection: close` is sent if shutdown was requested. This can be tested by ``` bitcoind -regtest nc localhost 18443 POST / HTTP/1.1 Authorization: Basic ... Content-Type: application/json Content-Length: 44 {"jsonrpc": "2.0","method":"stop","id":123} ``` Summing up, this PR: - removes explicit event loop exit — event loop exits once there are no active or pending events - changes the moment the listening sockets are removed — explained above - sends header `Connection: close` on active requests when shutdown was requested which is relevant when it's a persistent connection (default in HTTP 1.1) — libevent is aware of this header and closes the connection gracefully - removes event loop explicit break after 2 seconds timeout Tree-SHA512: 4dac1e86abe388697c1e2dedbf31fb36a394cfafe5e64eadbf6ed01d829542785a8c3b91d1ab680d3f03f912d14fc87176428041141441d25dcb6c98a1e069d8		2018-12-06 17:43:07 +01:00
..
functional	Merge #14670 : http: Fix HTTP server shutdown	2018-12-06 17:43:07 +01:00
lint	Merge #14831 : Scripts and tools: Use #!/usr/bin/env bash instead of #!/bin/bash.	2018-12-06 15:47:44 +01:00
sanitizer_suppressions	Move CheckBlock() call to critical section	2018-11-30 12:40:57 +02:00
util	Properly generate salt in rpcauth.py, update tests	2018-11-17 01:26:49 -08:00
config.ini.in	configure: Make it possible to build only one of bitcoin-cli or bitcoin-tx	2018-09-13 11:48:23 +00:00
README.md	docs: Document lint tests	2018-10-20 13:09:05 +08:00

README.md

This directory contains integration tests that test bitcoind and its utilities in their entirety. It does not contain unit tests, which can be found in /src/test, /src/wallet/test, etc.

This directory contains the following sets of tests:

functional which test the functionality of bitcoind and bitcoin-qt by interacting with them through the RPC and P2P interfaces.
util which tests the bitcoin utilities, currently only bitcoin-tx.
lint which perform various static analysis checks.

The util tests are run as part of make check target. The functional tests and lint scripts are run by the travis continuous build process whenever a pull request is opened. All sets of tests can also be run locally.

Running tests locally

Build for your system first. Be sure to enable wallet, utils and daemon when you configure. Tests will not run otherwise.

Functional tests

Dependencies

The ZMQ functional test requires a python ZMQ library. To install it:

on Unix, run sudo apt-get install python3-zmq
on mac OS, run pip3 install pyzmq

Running the tests

Individual tests can be run by directly calling the test script, e.g.:

test/functional/feature_rbf.py

or can be run through the test_runner harness, eg:

test/functional/test_runner.py feature_rbf.py

You can run any combination (incl. duplicates) of tests by calling:

test/functional/test_runner.py <testname1> <testname2> <testname3> ...

Run the regression test suite with:

test/functional/test_runner.py

Run all possible tests with

test/functional/test_runner.py --extended

By default, up to 4 tests will be run in parallel by test_runner. To specify how many jobs to run, append --jobs=n

The individual tests and the test_runner harness have many command-line options. Run test_runner.py -h to see them all.

Troubleshooting and debugging test failures

Resource contention

The P2P and RPC ports used by the bitcoind nodes-under-test are chosen to make conflicts with other processes unlikely. However, if there is another bitcoind process running on the system (perhaps from a previous test which hasn't successfully killed all its bitcoind nodes), then there may be a port conflict which will cause the test to fail. It is recommended that you run the tests on a system where no other bitcoind processes are running.

On linux, the test_framework will warn if there is another bitcoind process running when the tests are started.

If there are zombie bitcoind processes after test failure, you can kill them by running the following commands. Note that these commands will kill all bitcoind processes running on the system, so should not be used if any non-test bitcoind processes are being run.

killall bitcoind

pkill -9 bitcoind

Data directory cache

A pre-mined blockchain with 200 blocks is generated the first time a functional test is run and is stored in test/cache. This speeds up test startup times since new blockchains don't need to be generated for each test. However, the cache may get into a bad state, in which case tests will fail. If this happens, remove the cache directory (and make sure bitcoind processes are stopped as above):

rm -rf cache
killall bitcoind

Test logging

The tests contain logging at different levels (debug, info, warning, etc). By default:

when run through the test_runner harness, all logs are written to test_framework.log and no logs are output to the console.
when run directly, all logs are written to test_framework.log and INFO level and above are output to the console.
when run on Travis, no logs are output to the console. However, if a test fails, the test_framework.log and bitcoind debug.logs will all be dumped to the console to help troubleshooting.

To change the level of logs output to the console, use the -l command line argument.

test_framework.log and bitcoind debug.logs can be combined into a single aggregate log by running the combine_logs.py script. The output can be plain text, colorized text or html. For example:

combine_logs.py -c <test data directory> | less -r

will pipe the colorized logs from the test into less.

Use --tracerpc to trace out all the RPC calls and responses to the console. For some tests (eg any that use submitblock to submit a full block over RPC), this can result in a lot of screen output.

By default, the test data directory will be deleted after a successful run. Use --nocleanup to leave the test data directory intact. The test data directory is never deleted after a failed test.

Attaching a debugger

A python debugger can be attached to tests at any point. Just add the line:

import pdb; pdb.set_trace()

anywhere in the test. You will then be able to inspect variables, as well as call methods that interact with the bitcoind nodes-under-test.

If further introspection of the bitcoind instances themselves becomes necessary, this can be accomplished by first setting a pdb breakpoint at an appropriate location, running the test to that point, then using gdb to attach to the process and debug.

For instance, to attach to self.node[1] during a run:

2017-06-27 14:13:56.686000 TestFramework (INFO): Initializing test directory /tmp/user/1000/testo9vsdjo3

use the directory path to get the pid from the pid file:

cat /tmp/user/1000/testo9vsdjo3/node1/regtest/bitcoind.pid
gdb /home/example/bitcoind <pid>

Note: gdb attach step may require sudo

Util tests

Util tests can be run locally by running test/util/bitcoin-util-test.py. Use the -v option for verbose output.

Lint tests

Dependencies

The lint tests require codespell and flake8. To install: pip3 install codespell flake8.

Running the tests

Individual tests can be run by directly calling the test script, e.g.:

test/lint/lint-filenames.sh

You can run all the shell-based lint tests by running:

test/lint/lint-all.sh

Writing functional tests

You are encouraged to write functional tests for new or existing features. Further information about the functional test framework and individual tests is found in test/functional.