Upstream issues with boto in Duplicity

[elsiehupp]

For a couple weeks I've been getting this message from Mail-in-a-Box every time it runs a backup.

WARNING: Using boto3 >= 1,36.0 with non-amazon s3 services may result in checksum errors. a workaround is to set the following env vars

    export AWS_REQUEST_CHECKSUM_CALCULATION=when_required
    export AWS_RESPONSE_CHECKSUM_VALIDATION=when_required

see https://gitlab.com/duplicity/duplicity/-/issues/870 for details.

Note that, yes, I am in fact using a non-Amazon service for object storage.

I didn't find a relevant Issue here, but I found someone reporting the same message on the MiaB Discourse, so I initially responded there.

As best I can tell, the solution to this problem is to add the environmental variables to /etc/environment. One way to do this is with nano.

First, ssh into your Box. Then:

sudo nano /etc/environment

In nano, scroll down one line, paste the following:

export AWS_REQUEST_CHECKSUM_CALCULATION=when_required
export AWS_RESPONSE_CHECKSUM_VALIDATION=when_required

(You may want to add an extra line break at the end, hence the blank line in the blockquote.)

Next, type Ctrl-O to save the file, then Ctrl-X to exit nano.

The environmental variables should take hold the next time the server is rebooted (I think?), so if you want to make sure they've taken hold, you can manually force-restart with the following:

sudo reboot

Note that there are other, more elegant ways to add these environmental variables, such as with echo and >>, but I'm not sure how to put the sudo in the correct place for that, hence just using nano.

After rebooting, you can ssh in again check the variables with echo:

echo "$AWS_REQUEST_CHECKSUM_CALCULATION"

Should print when_required, and

echo "$AWS_RESPONSE_CHECKSUM_VALIDATION"

Should also print when_required.

Anyway, I feel like probably these environmental variables could be set by the MiaB installer without any detrimental side effects, and adding this action to the MiaB installer would probably count as a "fix" for this Issue I am creating.

[elsiehupp]

After 24h and another daily backup cycle I got another email with the same error, so the fix I proposed may or may not have actually worked… 😬

[dms00]

I may be wrong, but to my knowledge /etc/environment is not loaded into the environment by cron, and I could find no docs suggesting otherwise.
However, I believe you can add environment vars to the cron file itself. In this case, I believe you would want to change the file, /etc/cron.d/mailinabox-nightly because that's what runs the daily backup.

So, try adding the following 2 lines to this cron file above the cron command:

AWS_REQUEST_CHECKSUM_CALCULATION=when_required
AWS_RESPONSE_CHECKSUM_VALIDATION=when_required

This is a custom configuration so if it does work, then you may need to remember to reapply the change after upgrades, because it may get overwritten.

[elsiehupp]

@dms00 thanks for the suggestion!

[mastazi]

Wait a second, the env variables are already there #2490 I remember I reviewed this PR into mailinabox last year. Was there a regression @JoshData ?

@elsiehupp @dms00 did you actually stop receiving the warning message after adding the environment variables? And what Linux user's crontab did you add them to?

[mastazi]

Update @elsiehupp @dms00 @JoshData based on https://gitlab.com/duplicity/duplicity/-/work_items/870#note_2948098348 it seems they are triggering the warning on the URL of the bucket, so we will continue receiving the warning no matter what the env variables are (and those are already fixed by #2490 ). @JoshData I think this particular message should be filtered out because it's very annoying for us users who will receive the warning email every day until the end of time.