timkay / aws

Easy command line access to Amazon EC2, S3, SQS, ELB, and SDB
http://timkay.com/aws/
427 stars 126 forks source link

SQS send returns error code 0 on error #85

Open nibynool opened 10 years ago

nibynool commented 10 years ago

Hi again Tim,

Thanks for the speedy help last time, I've found another small issue and once again my lack of Perl skills is letting me down.

I have defined a bash function as follows

function notifyQueue {
    ./aws --silent --simple --region="${REGION}" send-message "${1}" -message "${2}" || { echo $?; echo "{\"error\":\"Could not notify queue\",\"id\":\"${POST_ID}\"}"; exit 1; }
    echo $?
}

This works fine the majority of the time, but every so often I get the following response from your script

+--------+-----------------------+----------------------------------------------------------------------------------+
|  Type  |         Code          |                             Message                                              |
+--------+-----------------------+----------------------------------------------------------------------------------+
| Sender | SignatureDoesNotMatch | The request signature we calculated does not match the signature you provided... |
+--------+-----------------------+----------------------------------------------------------------------------------+

When this happens I still get an error code of 0 returned.

I've noticed that in your documentation is is mentioned that you don't always return a non-0 value on error, but can you give me some pointers as to where I would need to detect this to get a non-0 value.

My use case is that after processing a message I need to send a message to a second queue, when that message fails I don't want to delete the current working message.

Thanks again.

nibynool commented 10 years ago

Ok, I think I found the issue :)

Starting at line 2520 you have the following code:

    else
    {
    # If there is an error response and we haven't yet set an error return, do it now
    if ( $result =~ /<ErrorResponse/ && !$ExitCode)
    {
        $ExitCode = 1;
    }

    print xml2tab($result) || xmlpp($result);
    }

    exit $exit_code;

I think the references to $ExitCode should in fact be $exit_code

timkay commented 10 years ago

Is your point that the $ExitCode variable is not used? This code was added by a merge request from another user, related to RDS.

Why do you think that code relates to the error message you occasionally see?

nibynool commented 10 years ago

Hi Tim, sorry if I wasn't clear.

It doesn't relate to the cause of the error, but does relate to the reporting of the error in the script's exit code. I tried changing the two $ExitCode occurrences to $exit_code (which appears 10 times in the code) and the script now returns an error code of 1 when there is an error response from AWS.

nibynool commented 10 years ago

As a side note, I think the cause of the error is something I'm doing, because it consistently happens on one particular message.... Debugging the cause is 100% my problem though :)

timkay commented 10 years ago

It is possible that there is a bug in the signing code, even though the code has been used by a zillion people for many years. This sort of issue has been reported very infrequently, and I haven't been able to reproduce it. If there is such a bug, it depends on specific keys. People are usually unwilling to share their keys with me for obvious reasons. I suggest the following work around: Using IAM, create a new set of keys that have the permissions you need to access SQS. If you are able to reproduce the issue, then you could update the keys have access to a specific test queue and then share the credentials with me (via timkay@not.com, not via github). At that point, I could take a look. It is also possible that the new keys do not exhibit the faulty behavior, in which case you have your work around.