Dear fellow Mail Operators,
we run a SaaS solution hosting web forums for our customers. We send
transactional email and opt-in notifications about new activities within
a community which by their nature contain some user-generated content.
During a regular check in Google's Postmaster Tools we noticed that
Gmail reports a Spam Rate of 27.3% for a single day in May which is
concerning. Our Postfix logs report nothing out of the usual, apart from
a single email being rejected as 'UnsolicitedRateLimitError', but being
delivered 5 seconds later into another of Gmail's MXs the *day before*
this spike in spam reports.
We also checked the Feedback Loop dashboard, in an attempt to at least
identify the customer in question, but that dashboard shows nothing and
it never did when we checked it in the past. It's showing blue dots at
the very bottom for the days where we reach enough volume for other
dashboards to show data, including the day in question. But clicking
that dot shows "Keine gekennzeichneten Bezeichner" ("no marked
identifiers").
This makes me wonder if we missed anything while setting up the support
for Gmail's Feedback Loop or if this is usual behavior due to our lowish
volume (less than 100 emails in Gmails direction on the day in question).
Mail is being sent with a 'MAIL FROM:<[email protected]>'
with the 'From:' containing an email address of the customer's custom
domain.
We're DKIM signing the emails using a key in the 'bounce.woltlab.cloud'
domain and add a 'Feedback-ID: customer_id:WCloud' header to all emails,
in an attempt to uniquely identify the customer in cases of spam reports.
We set up SPF records for the bounce.woltlab.cloud domain and also set
SPF records for the customer's domain, because the customer sets up the
domain using a CNAME record.
We set up both the bounce.woltlab.cloud and woltlab.cloud domains at
postmaster.google.com.
Checking https://support.google.com/mail/answer/6254652?hl=en-GB we
appear to tick all boxes:
- Our emails contain exactly one Feedback-ID header with a customer
identifier and a stable SenderID of 'WCloud':
> Feedback-ID: c_1234578:WCloud
The documentation is not clear regarding the use of optional fields,
though. Would we need
> Feedback-ID: ::c_1234578:WCloud
instead?
- The Feedback-ID header is DKIM signed by the domain registered in the
dashboard:
DKIM-Signature: v=1; a=rsa-sha256; c=relaxed/simple;
d=bounce.woltlab.cloud; [email protected]; q=dns/txt; s=wcloud;
t=1622622079; h=from : reply-to : to : subject : date : message-id :
list-id : list-unsubscribe : list-unsubscribe-post : mime-version :
content-type : feedback-id : from : to : subject : date : message-id :
feedback-id; bh=YOZK9mUAdTkOR1zUM4JWklWhBmsqy1Wau9HMb4sOvXI=;
b=dtvd3fXJnEUEGrSUu4z8sY2kx5nkw5tiS8zKS/Se6YZUgtAyVcWb4Dg+Ze/AqFeWr5rvD
q2W9+u9iSonz5yoV1e/X25cwyPhr29063KzZPJTQOtLD5Kcosz/U4Ur3YTA/YWIeXg0Afo3
GhsqucX/g3qpq2Hs28lR2zzqg3Ek+FntafocHjOeKBQDazs4nG4cMX6j3R9TPTFniG5jEdo
oMW+ErlNTzvEL0Z0mVjPI4rfTprdkjyLXZPVv9h+tlHCYir5meVQ1RTpImZKWtaXqP8UjVv
z4TDqqI8FoFS9HPGhNfPGbwfrktsO/7kj5FBSDSQdbrWLPKA4HhUJ3qbzUAA==
- The signing domain has SPF records set up.
- Our outgoing mailserver has a forward confirmed reverse DNS within the
signing domain.
The only thing we might be lacking is:
For a given day’s traffic, FBL reports are generated only if a given identifier
is present in a certain volume of mails as well as in distinct user spam
reports.
As the other dashboards, specifically the Spam Report dashboard, show
data we would suspect that we reached this minimum volume, but of course
we can't be sure there.
Does anyone of you have practical experience with Google's feedback loop
mechanism and might be able to identify if we are doing anything wrong
or if it's just the low volume?
Best regards
Tim Düsterhus
Postmaster WoltLab GmbH
--
WoltLab GmbH
Nedlitzer Str. 27B
14469 Potsdam
Tel.: +49 331 96784338
[email protected]
www.woltlab.com
Managing director:
Marcel Werk
AG Potsdam HRB 26795 P
_______________________________________________
mailop mailing list
[email protected]
https://list.mailop.org/listinfo/mailop
