Submitted by lvsys on Wed, 11/21/2012 - 09:12
Hello,
Our backups to Amazon S3 are failing (it's been 3 nights in a row) with the following message:
Failed to list S3 buckets : The difference between the request time and the current time is too large.
What is happening? How can we solve this?
Best Regards
MD
Status:
Active
Comments
Submitted by andreychek on Wed, 11/21/2012 - 09:21 Comment #1
Howdy -- Amazon requires the system time on your server to be correct.
If you run the command "date", does it list the correct time and timezone?
Submitted by lvsys on Wed, 11/21/2012 - 09:47 Comment #2
Ha. I'll be darned. The date was off by 15 mn. How much of a time delay does Amazon tolerate?
MD
Submitted by andreychek on Wed, 11/21/2012 - 10:00 Comment #3
Hmm, I'm not entirely certain, but in digging around a bit, I saw one reference to them having a tolerance of up to 15 minutes.
Which sounds like you're pretty close to that.
If you don't have it installed now, you could always install the ntp package, would should assist in keeping your system sync'd with the atomic clock.
Submitted by Locutus on Wed, 11/21/2012 - 12:42 Comment #4
I very much agree with Eric in terms of ntp usefulness. :)
I had some trouble with incorrect clock recently on my VMware virtual machines, and using the VMware Tools to sync time with the host made it even worse.
I switched to ntp for all my VMware needs meanwhile.
ntp has the big advantage that it doesn't adjust the clock "abruptly", like after a day it notices that the clock is off by 5 seconds and does a 5 second jump, but constantly runs in the background and does "micro adjustments", which don't add up to more than a second. At least that's what I understood how it's supposed to work. :)
There are services, e.g. Dovecot, which react very unpleasantly especially if the clock jumps backwards. ntp helps a lot here.