Strange exit after updated to node 8.0.0 #13325
Comments
mscdex
added
async_hooks
|
Any chance you could provide a core dump, or code to reproduce? |
|
right now not... not so easy but will try in a couple of hours |
|
Ok seems I have a core dump 1.49 GB and 264 MB zipped |
|
Wherever you want – if you don’t want to share it with everyone (i.e. because your program contains non-public code or data), you can also send it or a link to it to my email address that’s listed in the readme (I’ll take a look at that in a few hours then). |
|
Thanks a lot... I've sent you an email. |
|
Hmm … maybe somebody from @nodejs/platform-macos would be able to take a look, too? Right now I don’t have the tooling to deal with macos core dumps beyond pretty trivial things. A full disassembly of the 8.0.0 macos executable might be helpful as well; the stack trace as it is shown in the original issue should not really be possible, because I don’t see any way for |
|
@adrai Where did you install your node binary from? |
|
I have got the coredump from @addaleax (who got this from @adrai) and opened it. I am not familiar with the async hooks enough to know why the assertion failed but there could be an overflow which caused See the stacktrace |
|
@joyeecheung If you're feeling adventurous give https://github.com/nodejs/llnode a try:
|
|
@bnoordhuis Yeah I've been using it, frame |
|
Thanks, now we now it's the second argument on this line: https://github.com/nodejs/node/blob/v8.0.0/lib/internal/process/next_tick.js#L145 cc @AndreasMadsen @trevnorris (we don't have a async-hooks group yet, right?) |
|
@adrai could I get you to add this code to the beginning of your script (perhaps using the const async_hooks = require('async_hooks');
async_hooks.createHook({
init(id, type, triggerId, resource) {
if (triggerId < 0) {
process._rawDebug('init', {id, type, triggerId});
Error.stackTraceLimit = Infinity;
throw new Error('bad trigger id');
}
}
}).enable(); |
|
@AndreasMadsen Good idea. I would also turn the |
|
@addaleax Sure. const async_hooks = require('async_hooks');
async_hooks.createHook({
init(id, type, triggerId, resource) {
if (triggerId < 0 || id < 0) {
process._rawDebug('init', {id, type, triggerId});
Error.stackTraceLimit = Infinity;
throw new Error('bad trigger id');
}
}
}).enable(); |
|
I run to my notebook... :-) |
|
Is this ok? |
|
@adrai Yes, I think so, but unfortunately the stack trace is truncated – if you set |
|
|
I can reproduce it like this: const restify = require('restify');
const jsonTargetClient = restify.createJSONClient({
url: 'http://localhost:4999',
version: '*'
});
jsonTargetClient.basicAuth('wrong', 'creds');
setInterval(() => {
console.log('send the request...');
jsonTargetClient.post({
path: '/the/path'
}, { a: 'bit of data' }, (err) => {
console.log('responded...');
console.error(err);
});
}, 1000);the output is: send the request...
responded...
{ UnauthorizedError
at ClientRequest.onResponse (/Users/adrai/Projects/iot-cloud/service-out/node_modules/restify/lib/clients/http_client.js:169:26)
at Object.onceWrapper (events.js:312:19)
at emitOne (events.js:115:13)
at ClientRequest.emit (events.js:210:7)
at HTTPParser.parserOnIncomingClient [as onIncoming] (_http_client.js:563:21)
at HTTPParser.parserOnHeadersComplete (_http_common.js:116:23)
at Socket.socketOnData (_http_client.js:452:20)
at emitOne (events.js:115:13)
at Socket.emit (events.js:210:7)
at addChunk (_stream_readable.js:250:12)
at readableAddChunk (_stream_readable.js:237:11)
at Socket.Readable.push (_stream_readable.js:195:10)
at TCP.onread (net.js:586:20)
jse_shortmsg: '',
jse_info: {},
message: 'Unauthorized',
statusCode: 401,
body: {} }
send the request...
async_hooks.js:322
throw new RangeError('triggerId must be an unsigned integer');
^
RangeError: triggerId must be an unsigned integer
at emitInitS (async_hooks.js:322:11)
at setupInit (internal/process/next_tick.js:225:7)
at internalNextTick (internal/process/next_tick.js:269:5)
at ClientRequest._writeRaw (_http_outgoing.js:279:9)
at ClientRequest._send (_http_outgoing.js:255:15)
at ClientRequest.end (_http_outgoing.js:780:16)
at /Users/adrai/Projects/iot-cloud/service-out/node_modules/restify/lib/clients/string_client.js:108:17
at f (/Users/adrai/Projects/iot-cloud/service-out/node_modules/once/once.js:25:25)
at FunctionCall.doCallback_ (/Users/adrai/Projects/iot-cloud/service-out/node_modules/backoff/lib/function_call.js:162:20)
at FunctionCall.handleFunctionCallback_ (/Users/adrai/Projects/iot-cloud/service-out/node_modules/backoff/lib/function_call.js:181:14)
at f (/Users/adrai/Projects/iot-cloud/service-out/node_modules/once/once.js:25:25)
at ClientRequest.onSocket (/Users/adrai/Projects/iot-cloud/service-out/node_modules/restify/lib/clients/http_client.js:225:13)
at Object.onceWrapper (events.js:312:19)
at emitOne (events.js:120:20)
at ClientRequest.emit (events.js:210:7)
at tickOnSocket (_http_client.js:666:7)
at onSocketNT (_http_client.js:682:5)
at _combinedTickCallback (internal/process/next_tick.js:102:11)
at process._tickCallback (internal/process/next_tick.js:161:9)Perhaps @retrohacker has an idea? |
|
With the following script: const http = require('http');
const restify = require('restify');
const stringClient = restify.createStringClient({
url: 'http://localhost:4999',
version: '*'
});
http.createServer(function (req, res) {
res.writeHead(401, {'Content-Type': 'text/plain'});
res.end('error');
}).listen(4999, function () {
setInterval(() => {
console.log('send the request...');
stringClient.post({
path: '/the/path'
}, 'data', () => {
console.log('responded...');
});
}, 100);
});I can reproduce the crash with Node 8.0.0 but NOT with |
|
correlated with #13237 ? |
|
Weird: I am unable to reproduce with a fresh build of |
|
I can reproduce this, too (only on 8.0.0 as well, not master).
I don’t think so – that should be fixed in 8.0.0, and only affects you if you use async_hooks actively. |
|
Hmmm, the binary on nodejs.org aborts for me but a custom build of |
|
This still is an issue on So, as far as I can tell, the problem |
|
I hope there is someone who knows what would be the best for everyone... |
|
We had a suspicion this could happen with a custom |
|
@refack Sorry, my initial comment above was wrong, this is about not using a custom |
|
So I see a race on the kept alive socket between The Socket does not pass through the constructor to reset it's Somewhere we need to do another |
Under very specific circumstances the `http` implementation could be brought to crash, because the Agent did not re-assign the async id field properly after setting up a socket for reuse. Fixes: nodejs#13325
|
Any update? ETA on next 8.x release ;-) |
|
@adrai The fix should be ready to be merged tomorrow (we have a minimum waiting time for that kind of thing), and I think we’re planning for a release next week to fix up some things after the 8.0.0 release. |
|
thx a lot I use node.js since early 0.4 versions (I think 2011) but I'm always surprised by the awesome node.js community! |
Under very specific circumstances the `http` implementation could be brought to crash, because the Agent did not re-assign the async id field properly after setting up a socket for reuse. Fixes: #13325 PR-URL: #13348 Reviewed-By: Refael Ackermann <refack@gmail.com> Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Andreas Madsen <amwebdk@gmail.com>
|
Note this fix did not fix the issue in Haraka (which isn't using http). |
|
@baudehlo can you open an issue and show us the error? (also if it appears to be exactly the same.) Since you are not using |
|
Will do. |
|
@AndreasMadsen - I am running into this in my backend servers on all 8.x versions to included 8.1.4. Core dump available in s3 bucket |
|
@d3viant0ne please open a new issue. It is the same assert but most likely a different issue. |
|
No problem, thanks :) |
|
Updating to Node v9.4 seems to have solved my problem. |
Right now I don't know exactly what is causing the problem.... (I guess posting a http request to some endpoints)
The process exits with this:
Any guess?
The text was updated successfully, but these errors were encountered: