Repository navigation
Try to seek for race conditions occurring in Actions only #445
Description
Activity
We can move this out of Backlog when we get some specific details about particular tests that fail sometimes but not always in Actions. When you experience that, post a description in a comment here and attach the downloaded console log from Actions.
- changed the title
[-]Try to seek for Race Conditions occurring in Actions only: masq/commu...[/-][+]Try to seek for Race Conditions occurring in Actions only[/+]on Feb 12, 2021 - changed the title
[-]Try to seek for Race Conditions occurring in Actions only[/-][+]Try to seek for race conditions occurring in Actions only[/+]on Feb 12, 2021 I got this despite this area of our code could not be affected by the worked card (GH-381). Happened after a break from seeing any strangely failing tests. When a new Rust version came.
thread 'main' panicked at 'ClientListenerThread could not be restarted: Timeout', D:\a\Node\Node\masq\src\communications\connection_manager.rs:455:23
stack backtrace:
0: 0x7ff74c24187e - std::backtrace_rs::backtrace::dbghelp::trace
test communications::connection_manager::tests::handles_outgoing_fire_and_forget_messages_to_dead_server ... FAILED
at /rustc/cb75ad5db02783e8b0222fee363c5f63f7e2cf5b/library\std\src....\backtrace\src\backtrace\dbghelp.rs:98
1: 0x7ff74c24187e - std::backtrace_rs::backtrace::trace_unsynchronized
at /rustc/cb75ad5db02783e8b0222fee363c5f63f7e2cf5b/library\std\src....\backtrace\src\backtrace\mod.rs:66
2: 0x7ff74c24187e - std::sys_common::backtrace::_print_fmt
at /rustc/cb75ad5db02783e8b0222fee363c5f63f7e2cf5b/library\std\src\sys_common\backtrace.rs:67communications::connection_manager::tests::handles_outgoing_fire_and_forget_messages_to_dead_server
One of the multinode tests failed: provided_and_consumed_services_are_recorded_in_databases.
logs_473.zipThe same test...Bert.
2_Build node on ubuntu-latest.txtWindows ...one of node unit tests. No relation to the card being coped (GH-373).
configurator_standard ... setting the gas price
windows:
tls_through_node_test(integration); I know this test because that's the one which permanently fails on my Win machine.MacOS: Node: unit_tests:
initialize_with_clandestine_port_produces_expected_clandestine_discriminator_factories_vectorThis is unfortunate. By all evidences, GitHub stopped to support zipped packages and these ours which are attached to the comments here are inaccessible by now. I'd be happy if I was wrong.
Now it accept zipped folders again...so...I don't know. But we lost the old ones.
handle_startup_and_shutdown_integration
This test was renovated for this card, MASQ-Project/MASQ-Node-issues#386, however, usually it passes, this time it didn't. Maybe RC.
3_Build node on windows-latest.txt
...
3_Build node on windows-latest.txtAnd another one:
22.6.21
3_Build node on windows-latest.txtHmmm...this is growing up fast
24.6.2021
3_Build node on windows-latest.txt
26.6.2021 twice
3_Build node on windows-latest.txt
3_Build node on windows-latest.txtIn MASQ-Project/MASQ-Node-issues#451, I was trying to fix this by adding a piece of logic to the code. It may happen that this error will evaporate then.
Nope, without a relief, the frequency of this failure may even have increased.
3_Build node on windows-latest.txt
27th of September
I tried to prevent this issue more reliably with a modification in fallback - connection manager. Card MASQ-Project/MASQ-Node-issues#305.4th October
Not fixed yet12 remaining items
blockchain::blockchain_interface::tests::blockchain_interface_non_clandestine_can_retrieve_eth_balance_of_a_wallet
2_Build node on ubuntu-latest.txtmultinode
multiple_stream_zero_hop_test
2_Build node on ubuntu-latest.txtpass_args_or_print_messages_work_under_fine_lock_for_eof
I guess this is a rare one. I suggest just to increase the multiplier in the assertion like from 3 to 4 or something. I know there is a couple of tests like this and this was the latest I added, maybe the chosen multiplier doesn't fit as well as for the other tests, I think its value could be a bit test specific.
tls_through_node_integration
This one is really rare in Actions I think, on the other hand we come to see it quite often on dev machines. Speaking mainly for myself (maybe occurred almost all the time) but I also saw it at Utkarsh.
The multinode test tls_end_to_end_routing_test hung for most of five hours. We don't know why, but here's the log.
graph_connects_but_does_not_over_connect
1_build.txt
2_Build node on ubuntu-latest.txthttp_through_node_integration
masq_erc20_contract_exists_on_polygon_mumbai_integration
Metadata
Metadata
Labels
Type
Projects
- StatusShow more project fields📋 Backlog
- StatusShow more project fieldsBacklog

I've experienced them more than little times in: Node/masq/communication/communication_manager
at more places in there.