Closed jhungerford closed 4 years ago
I am having the same problem after upgrading to aws-java-sdk-s3 version 1.11.769
Thanks for reporting the issue, taking a look
Hi @jhungerford, @will-amp, we have pushed out a fix, and it's available in 1.11.772
Could you try with the latest version? https://github.com/aws/aws-sdk-java/blob/master/CHANGELOG.md#111772-2020-04-29
Thanks - that was fast! I'll update our code to use 1.11.772
, and report back here after I've let it soak for a while.
I haven't seen any more clock skew request failures after upgrading to 1.11.772
- thanks again!
I work on a Java 11 service running on centos in EC2 that makes 10's of thousands of AWS API calls every hour. After upgrading the sdk from
1.11.759
to1.11.768
, roughly 0.1% of API requests fail with:Most come from the RDS and Redshift RI describe endpoints, but I've also seen the same issue when sending messages to SQS.
I'm not sure if this is related to #2304 - my understanding is that the clock sku correction is used to make the auth token more reliable, but I'm seeing a different stack trace than the one in that issue.
Current Behavior
Example stack traces:
Sending a message to SQS:
Describe RDS Reserved Instances:
Steps to Reproduce
Since such a low percentage of requests fail because of clock sku, I haven't been able to reliably reproduce this issue. The reported serverTime varies between requests, and I've seen the error from several different regions.
In millis since the epoch,
-61525806985000
is ~51AD, so another possibility that I'm considering is that AWS unearthed an ancient roman data center while opening the new region in Milan.Your Environment
1.11.768