aws-sdk-js-v3: S3.send sometimes leads to application exit with code 0

I open a PR for this: https://github.com/aws/aws-sdk-js-v3/pull/4492 @trivikr

I’ve opened a PR to fix this https://github.com/aws/aws-sdk-js-v3/pull/4492

Hey, we have probably the same problem.

const main = async function() {
    const dynamodb = DynamoDBDocumentClient.from(new DynamoDB({}));
    const s3 = new S3Client({});
    const config = {
        client: dynamodb,
        pageSize: 20,
    };
    const input = {
        TableName: TABLE_NAME,
    };
    const paginator = paginateScan(config, input);
    let count = 0;
    for await (const page of paginator) {
        count += page.Count!;
        //@ts-ignore
        for (const item of page.Items) {
            if (item.PK.startsWith('MY_KEY#')) {
                if (!item.S3Version) {
                    try {
                        console.log('getting object', item.PK, item.S3Key);
                        const result = await s3.send(new GetObjectCommand({
                            Bucket: BUCKET_NAME,
                            Key: item.S3Key,
                        }));
                        console.log('updating measurement object', item.PK, item.S3Key);
                        const input1 = {
                            TableName: TABLE_NAME,
                            Key: { PK: item.PK, SK: item.SK },
                            UpdateExpression: 'set S3Version = :v',
                            ExpressionAttributeValues: {
                                ':v': result.VersionId,
                            },
                        };
                        console.log(input1);
                        if (false) {
                            await dynamodb.send(new UpdateCommand(input1));
                        }
                        await sleep(500);
                    } catch (e) {
                        console.log('skipping becasue of error...');
                    }
                }
            }
        }
    }

When I iterate over the DynamoDB table without S3 interaction it just works as expected. Once S3 is involved it just silently fails in the middle of the process. That’s kind of unfortunate. node: v16.18.1

Hi @bes,

Thanks for providing me with the repo.

I refactored your code a bit and let the program run for about 5-6 minutes before stopping it. It managed to write a few thousand objects to s3 successfully without exiting:

I usually use unique Object bodies when testing because otherwise the Etag would be the same, and it would be harder to debug. So Something like:

Body: `hello ${Date.now()}`

The only time my application crashed was when I ran the request with an empty buffer like in your example:

Body: Buffer.from("", "utf-8")

The error I got was

node[42307]: ../src/node_http_parser.cc:589:static void node::(anonymous namespace)::Parser::Execute(const FunctionCallbackInfo<v8::Value> &): Assertion `parser->current_buffer_.IsEmpty()' failed.
 1: 0x1025c2e28 node::Abort() [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 2: 0x1025c2c68 node::AppendExceptionLine(node::Environment*, v8::Local<v8::Value>, v8::Local<v8::Message>, node::ErrorHandlingMode) [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 3: 0x1025dc004 node::(anonymous namespace)::Parser::Execute(v8::FunctionCallbackInfo<v8::Value> const&) [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 4: 0x102785274 v8::internal::FunctionCallbackArguments::Call(v8::internal::CallHandlerInfo) [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 5: 0x102784d58 v8::internal::MaybeHandle<v8::internal::Object> v8::internal::(anonymous namespace)::HandleApiCallHelper<false>(v8::internal::Isolate*, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::HeapObject>, v8::internal::Handle<v8::internal::FunctionTemplateInfo>, v8::internal::Handle<v8::internal::Object>, v8::internal::BuiltinArguments) [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 6: 0x1027845e4 v8::internal::Builtin_HandleApiCall(int, unsigned long*, v8::internal::Isolate*) [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 7: 0x102f57a2c Builtins_CEntry_Return1_DontSaveFPRegs_ArgvOnStack_BuiltinExit [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 8: 0x102ee1b98 Builtins_InterpreterEntryTrampoline [/Users/rvaknin/.nvm/versions/node/v18.0.0/bin/node]
 ...
 ...
 ...
 ...

And this went away when I commented out process.exit(0) which was my lead suspect to begin with. Did you try commenting that out?

Thanks, Ran~

I see from the stack trace that you are executing this with Node 18.0.0. In 18.6.0 a fix for this panic was introduced. So please make sure you use a more recent Node version.

Hi @bes,

I still have no idea what the issue is and why you are the only one experiencing this. I will discuss it again with the team and see some of the more senior peeps can take a deeper look.

Thanks, Ran~