Tell HN: Gmail rate limiting emails from AWS SES | Hacker News Reader

Tell HN: Gmail rate limiting emails from AWS SES

67 pointssaradhi87 comments

Unsure if this is the same for everyone but our notification emails from SES are being rate limited by Gmail.

Background:

One of my apps provides time-sensitive email notifications, for 60-90 days, to the subscribers(paid) - so it is critical to deliver emails. We have been using the same email provider and have been using AWS SES for a couple of years now. We have the SPF, and DKIM all verified. Yet, for the last 3 days, we are getting the below email

```

Sub: Delivery Status Notification (Failure)

Body: Our system has detected an unusual rate of<CRLF>421-4.7.28 unsolicited mail originating from your IP address. To protect our<CRLF>421-4.7.28 users from spam, mail sent from your IP address has been temporarily<CRLF>421-4.7.28 rate limited. Please visit<CRLF>421-4.7.28 https://support.google.com/mail/?p=UnsolicitedRateLimitError to<CRLF>421 4.7.28 review our Bulk Email Senders Guidelines.>

```

Troubleshooting till now: I have got the AWS tech support team confirming my email configuration has no issues. AWS team has informed "The current throttling is just that Gmail is seeing a lot of messages from the SES shared IP and is throttling the messages"

>>> Is anyone facing the same issue with AWS? or similar issues with other bulk email service providers?

>>> How do you deal with such issues in the future? Set up alternative email service providers.

>>> Is this the side effect of Gmail's dormant account deletion rolled out last week?

View on HN

Things like this are to be expected with shared IP addresses for mail services.

This sounds like something Amazon should figure out more than anything. There's probably something going on with a customer or theirs (misconfiguration, spammer, vulnerability exploited) that's triggering spam detection rules.

At least Google reports the issue back to you instead of silently dropping all of your email like many other mail servers do.

Very useful thread. I recently considered SES for a quick app and didn't discover the purpose of dedicated IPs. They're listed under add ons, but seem somewhat essential if the emails are important.

Checking out the pricing page, a dedicated IP costs $24.95 per month, which would be more than the actual cost of emails for many small-medium apps (e.g. 100K emails per month would be about $10).

Ooc, do other email providers have the same shared-IP problem? E.g. Mailgun, Postmark, Sendgrid?

Source: https://aws.amazon.com/ses/pricing/

do other email providers have the same shared-IP problem?

Yes, that is why they let people get a dedicated IP pools and FCrDNS using their corporation name for a price. It is not uncommon for the shared IP pools to get rate limits, temporary blocks, etc... and the corporation specific dedicated IP's are more likely to get through especially if they handle UCE complaints properly.

> Yes

I didn't spend a huge amount of time reading the docs (just enough to get up and running), but I didn't notice this about other providers either. Great to be aware of this for future reference.

Postmark spends A LOT of effort to keep their shared IPs perfect. Rather complex setup with verification. And lots of stuff in place to prevent anything spam related.

But everyone else doesn't really care much.

Dedicated can be an issue unless you have sufficient volume, unless I've been mislead in the past.

Sendgrid's cheapest plan with a dedicated IP is $89/mo.

I finally started sending anything from AWS SES straight to trash because the spam on their platform is persistent and they are non-responsive to spam reports. I was getting 30+ fraudulent/spam emails a day in just one inbox alone.

That's untrue. I can tell you that spam reports and bounces are monitored, and customers do have to go through validation of their workflows if they breach thresholds.

I had to help multiple customers who had access to SES shut off for valid use cases because they didn't handle events coming back from customers appropriately.

  I can tell you that spam reports and bounces are monitored

If they are, I saw no evidence of that (except for the initial auto-response) based on the numerous spam reports I manually submitted with headers, dates/times and body content.

If you work for SES, you all aren't acting swiftly enough to shut down egregious spam operators using your platform that are making other people's lives miserable.

I have never gotten a response to an SES spam report.

A quick search turns up 50 or so in my sent mail folder within the past year or two.

3 of the most recent 5 reports are all the same message content (likely the same sender, although possibly using different—maybe hacked—accounts).

By way of comparison, I've sent a similar volume of complaints to Twilio Sendgrid, received a response to every single complaint, and a random sampling of recent reports doesn't show any repeat message content.

Google doesn't even provide a way to report spam from Gmail users, which is ironic given the OP is complaining about Google's aggressive spam blocking. Spam from Gmail is comparable in volume to what I receive from SES and Sendgrid, and almost all of those messages are obvious phishing and 419 scams. (One would think one of the largest tech companies in the world would be able to implement basic filters to catch these, if they cared enough to do so.) Cf. spam from SES, which is mostly either cryptocurrency pump-and-dump, or semi-legit companies sending to spamtrap addresses that appear on publicly-indexed web pages, in WHOIS, etc. but which have been never used for real correspondence (and have obviously never opted in to any lists).

No response != nothing being done with the report

Sending the report to /dev/null would technically be doing something with the report though.

My problem is "no response to report" combined with "I continue to receive the same or substantially similar volume and types of emails from the same sender(s) from the same email provider weeks after submitting the reports" == "non-responsive email service operator".

Your experience of customers not handling events appropriately - and then having SES access revoked - does not negate or disprove alyandon's experience of getting no responses to spam reports.

You may have had a different experience, but that doesn't make their (differing) experience "untrue."

Not sure if this is an edge case or not. But getting approved for SES isn’t that simple. And then shut downs happen very quickly even for legitimate cases. AWS is very aggressive with SES.

Maybe they care now. They certainly did not appear to care when my mailbox was getting spammed into uselessness by their customers.

Google also silently drops mail in some cases.

I've never experienced that. When mail gets dropped, my experience is that you either receive a notice from the SMTP service or get a DMARC report that tells you about the percentage of dropped mail.

Please explain how you measured your email deliverability.

Which cases?

Impossible to know. But in my cases from any IP with low volume. I.e. I have a bunch of personal servers for 10+ years, there are rarely outgoing emails from them and google happily just black-hole emails from such low volume IPs.

I'd be surprised if this was accurate. A low-volume IP might get a temporary failure code 4xx, but that's not silent. If they accept your message they intend to deliver it, or bounce it. Perhaps your return path is defective.

Google used to do this for low volume emails I was literally sending to myself by authenticating with my google smtp account credentials. I could find no rhyme or reason to it. It would claim to accept emails and they would never show up in my inbox or in spam. Sometimes they would get delivered to spam. Sometimes to my inbox like expected.

It most often involved security log snippets which would include ip addresses and hostnames that were likely already on "is a known bad actor" list.

Bug on their end? Intentional thing to disrupt spammers? I don't know.

Well, I won't contradict your personal experience, but I will say that I worked on gmail delivery for six years and if during that time it had ever come to light that any message had been accepted and not ultimately disposed of in some way, that would have been P0 and dozens of people would have dropped everything in order to root-cause and remediate that.

> Well, I won't contradict your personal experience, but I will say that I worked on gmail delivery for six years and if during that time it had ever come to light that any message had been accepted and not ultimately disposed of in some way, that would have been P0 and dozens of people would have dropped everything in order to root-cause and remediate that.

How would you even know? Last I checked, I was unable to find any support path to Gmail where one could report email-blackholing-instances and get them investigated.

This is what I wrote several years ago at the point when I gave up trying to run my own email server:

> Over the course of 2 years that I was actively using my own server to send email from the same domain and same IP, there were periods of time when my emails were landing in Gmail’s inbox. And then there were periods of time when my emails were placed in spam, sometimes outright bounced, and sometimes received and dropped without even landing in the spam folder. You could never know what Gmail would do with your email.

https://www.attejuvonen.fi/dont-send-email-from-your-own-ser...

That's cool and all and I'll take your statement in good faith.

However, how do I as a user report that an email that was accepted by Google's own SMTP servers was not subsequently delivered to my Gmail inbox? I can't call or email support and it is a meme at this point that Google support is only accessible via social media or contacting a Google insider.

I only have the perspective from the edges of the organization, and from years ago, but I can’t recall investigating anything other than support tickets that came from account representatives. So you probably need a paid account. That said, these are as little as $1.99/month.

I am certain that if you have a case that reproduces even one in a million times the relevant teams will be highly attentive.

This was really only a problem 3 or 4 years ago. At some point in time, it just stopped happening and I hadn't really thought about it since then. I am a Google One subscriber so there are probably some support options open to me that I could avail myself to if the issue ever resurfaced that weren't available to me when I wasn't a subscriber.

Email sending is a largely efficient market, and SES is the cheapest sender.

Thus, the SES IP pools have the worst IP reputations among all the SMTP-based senders. I would never use them (or an ESP that sends with them) in a million years.

The reason why cheap = bad in email is because Spammers have the lowest conversion rates of all senders (their emails are always untargeted), so price is their number 1 consideration.

You can cheap out elsewhere in your stack. But never cheap out on email — especially if you’re not sending high volume enough for a dedicated IP.

Thanks

I never had to use SES or really Email for that matter in my own architecture.

Since we get so immersed in best practices and recommendations of good architecture pushed by these very cloud providers (AWS, Azure, etc) -- this kind of candid information is usually not available unless we burn our fingers and learn through own experience.

I wish there was more efficient way of such practical knowledge to be shared among practitioners. Current discussion forums etc work to an extent but just hope there was much more effective way to spread awsreness.

> best practices and recommendations of good architecture pushed by these very cloud providers

...are deliberately designed to mix several actual best pracices and a lot of marketing so that it's difficult to discern what is what.

This is bad advice.

SES is the best among the provider.

They treat spam very seriously, compare to other. You would need to maintain a baseline, failure to do so will get the account into soft-susspend or permanent suspended quickly.

AWS, being an engineering focus product, provider tooling and automated around bouncing, spam reporting handling. Having sophisicated tooling help user deal with that.

They strongly againtst and not suggest end-user to buy dedicated IP. Where as other providers always want user to pay more for "dedicated ip" to "get better delivery".

AWS has procedure, best practice to encounrage slowly warmup when sending mass volume. They had their own rate limit (14 messages per second by default), move the account out of sandbox, a good domain/sender verification.

They are the best among providers when it comes to email.

Sometimes they appear to be worst than others, but that is a specific case. The way email works will always have false positive. If user decided the email is spam(even it isn't) and keep reporting it may appear in a certain spam list.

Every SMTP service (Mailgun, Sendgrid, Postmark, etc.) does the same things AWS does to try to prevent spam and has the same features you mention (shared IPs to start, warmup on mass sending, developer tooling, etc.)

Email delivery is a commodity business at this point, hence why there's so many PE rollups in the space.

The fact of the matter is it only takes 1 bad send to end up on a blacklist, and SES is the cheapest so it will always be the most attractive target for spammers/fraudsters.

Am I correct that maybe you run a Saas built off SES and might have incentive to defend?

I run a competitor of SES so I won't defend them.

But the problem is that Mailgun/Sendgrid/Postmark and the like is less senstive to spam and ban. The ban on SES is way heavier and the spam rule is stricer.

Maigun/Sendgrid even suggest to use dedicated IP pool to improve delivery, where as AWS SES doesn't https://docs.aws.amazon.com/ses/latest/dg/dedicated-ip.html https://docs.sendgrid.com/ui/account-and-settings/dedicated-... You can see the tone there.

You are assuming that SES allows for spammers. In my experience, SES has pretty good controls , limits and policies to dissuade from spamming.

Literally all ESPs and SMTP services have super strict anti-spam policies. The problem is it only takes 1 bad send to end up on a blacklist.

The cheapest service will always get the most attempts from spammers. You can try to detect them before they send, but the only way you know for sure that they're a spammer, is after they've already sent spam using your platform.

I'm a bit surprised that the spammers even bother with it, since it costs money and the conversion rates are near zero. I had the impression that most spammers would rather take over hacked accounts -- not quite zero cost, but pretty low.

If SES is the worst, who is the best? By implication, simply the most expensive providers?

Not necessarily (although, hard to imagine there's even 1 spammer willing to spend $100,000/yr on some enterprisey crap like Salesforce Pardot or Marketo).

But I'd just go with whatever sender isn't the cheapest and has the founders still intimately involved in the product. Stopping spammers is ultimately still a human game. Postmark used to be my go-to, but they sold, and I'd bet my entire salary that within a few years that service will end up like all the others that have been bought-and-sold.

It sounds like you could benefit from having your own dedicated IP pool in SES.

https://docs.aws.amazon.com/ses/latest/dg/dedicated-ip.html

Amazon SES is not a good choice for sending critical email notifications. Their 'global suppression list' [1] has caused no end of headaches for me and my clients.

If you and I are both using SES to send to the same person, and my message results in a hard bounce, then your messages to that person will start to silently fail.

[1] https://docs.aws.amazon.com/ses/latest/dg/sending-email-glob...

You can now use a account-level suppression list to override the global suppression list though.

> If an address is on the global suppression list, but not on your account level suppression list (which means you want to send to it), and you do send to it, Amazon SES will still attempt delivery, but if it bounces, the bounce will affect your own reputation

[1] https://docs.aws.amazon.com/ses/latest/dg/lists-and-subscrip...

Interesting. I've worked with vendors (Including VMWare/Carbon Black) that told me they couldn't override the global list. Maybe the product is evolving to address these (past?) flaws

Would you happen to know of better alternatives for sending critical email notifications? We're just now working on moving away from Mailgun into SES. This thread is making me reconsider.

"Critical" and "email" don't generally go together.

To echo the other reply, I don't rely on email for critical comms.

If I had to rely on email for critical comms, I would use a product that lets me see the SMTP logs, and alert on failures. I would have some backup provider that I could then quickly shift to

But I'm still an on-prem email administrator, so my entire paradigm is pretty much obsolete. I may not be the best source of info in this space

Thanks you all for comments. I have made a decision to subscribed to dedicated IPs (credits: @slau).

The differentiating factor between our current AWS SES plan and the competitors (mentioned in the comments) is having a dedicated IP. With our current volume, none of the competitors are anyway near AWS SES costs. So, moving to a dedicated IPs thats cost 25$ extra not only solves our issue, but also no change in code/infrastructure.

Just make sure you have sufficient traffic to warrant a dedicated IP. An unknown IP suddenly sending a burst of emails is going to get soft-blocked very quickly. You need to build up your reputation, and you need to slowly increase how much you send.

The managed IP is an option, although I’ve never used it.

When I was the VPoE at Dixa, we switched over to SES, and we had 3 dedicated IPs, and for our volume back then (a few thousand emails a day), this worked very well. I don’t know if they ever hit scaling issues after I left.

Hey everyone, CEO of MailChannels here. I've been following this thread with interest, as we've also observed similar challenges in the email delivery space. Well, to be honest, the ground game is constantly changing in this space as everyone has scaled.

Email delivery is inherently complex due to the various factors that contribute to deliverability, including IP reputation, domain reputation, content filtering, and recipient engagement. Shared IP pools can indeed be challenging because of the "bad neighbor" effect, where one sender's bad behavior can affect the reputation of all senders using that IP.

However, shared pools can also prove advantageous because it's harder for a receiver to block your IP if tons of email comes from it from a wide variety of senders. Receivers are trying to reduce collateral damage while protecting their users from spam and phishing - this is literally the reward model feeding their machine learning models. If your email travels alongside millions of other emails that are mostly received well, that IP will not be blocked; whereas, if you send email from your own IP, it doesn't take much for a receiver to pull the trigger and block you since there is very little consequence other than blocking your traffic.

Not that anyone here asked, but if you want a "best practice", try multiple different services and approaches and find the one that works best for you. There is no perfect email sending service for all senders and as mentioned above, the ground game is changing all the time.

This is typically not a big deal, as explained in the message, it is a temporary countermeasure. It'll resolve itself as long as you really aren't spamming.

Though Gmail responds citing your IP, Gmail and all other large email services don't use IP filtering. Just about all email service providers use domain reputation, since IPs are ephemeral.

If you are sending transactional emails that your customers have agreed to, then your domain (!= ip) rating will improve over time and there will be less countermeasures, regardless of which IP you use to send.

> Is anyone facing the same issue with AWS? or similar issues with other bulk email service providers?

This is just Gmail doing it's thing (the right thing, in my opinion, contrary to most HN sentiment). It is independent of which sender you use.

> How do you deal with such issues in the future? Set up alternative email service providers.

Use DMARC reporting to verify that all your email is sent with DKIM alignment, to make sure that you aren't causing the problem. This is independent of email service provider.

But as explained, you are being rate limited, not blocked. Email will be delivered, it'll just take longer. You state that you have a 60-90 day margin for delivery, so I wouldn't worry about it too much.

> Is this the side effect of Gmail's dormant account deletion rolled out last week?

No.

Not entirely true. There is a reason Mailchimp owns a rather large block of IP addresses.

https://ipinfo.io/AS14782

> Not entirely true

OK you are right, Gmail may use IP-based rating, but it only does that if there isn't sufficient proof (in form of DKIM signatures) that the email is sent on behalf of the domain. If the email is DKIM aligned, then domain rating is used. I just didn't want to go that far into detail in my post.

> There is a reason Mailchimp owns a rather large block of IP addresses.

The reason is mostly for supporting smaller/legacy/self-hosted email services that do rely on IP reputation. Since Mailchimp won't allow customers to bulk email unless they have DKIM set up correctly, they don't have to worry as much about IP spreading with the major email service providers.

I don't think this is accurate. Address reputation is an important signal regardless of DKIM.

I'm currently using PostmarkApp (from before the acquisition) but I've looked longingly at SES for years. My traffic is very bursty and so ~8 months out of the year I pay the monthly cost and send 1-2 emails if that and then the other 3 months I send close to my plan max. I'd love to switch to a pay-per-use provider but stories like this scare me. I've already dealt with deliverability issues (iCloud randomly deciding that unless you can receive emails, have the MX records in place, they will block you. This was for transactional/login/notification emails), since email is the login method for my sites it's rather important that it works. To PostmarkApp's credit they helped me to track down why the emails were bouncing to iCloud, I doubt I would have gotten the same support from AWS (I'm too small of a fish).

I'd love to hear what other people are using to send transactional emails (no marketing). Ideally I'd find a provider that could "scale down to $0".

If you are not in the business of selling email delivery, you should be buying email delivery from a company who is in that business. It's extremely hard to get emails delivered and it's even harder for a small company. For your use case you could probably use Postmark and get good delivery with that.

Unfortunately Postmark was recently acquired by a so-called “campaign management” company and they have already started sending unsolicited commercial mail to unauthorised recipients, and when I pushed back and complained they became downright hostile and confrontational and told us to unsubscribe - from a marketing list that we hadn’t opted into, and that was operated via their new parent company. It was like a conversation from the dark ages of UCE where the spammer says “just opt out mate”. So Postmark are dead to me now.

Holy crap! I had no idea. Do you mean they sent unsolicited commercial mail to registered users of Postmark, or do you mean they sent to email addresses they collected as part of their operations (emails that users of Postmark sent emails to)?

The former, but not even the primary details of a registered account; they subscribed the emergency contact email addresses, which for us included one of my board members, who then contacted me saying “why am I getting this noise?”

It was neither an emergency nor a transactional message, making it exactly the kind of behaviour they don’t permit from their own customers. Doubling down with the instruction to unsubscribe oneself could only be interpreted as a “fuck you”, and coming from an email company claiming such a moral high ground as Postmark formerly occupied, a sensational own goal.

Using SES is buying email delivery from a company who is in that business.

Disagree here. The sendgrid, mailchimp, postmark of the world are in the business of sending emails. They have their dedicated IPs and handle deliverability, anti spam, and whatnot with email providers like Gmail.

SES is an email sending infrastructure tool. That's not the same.

IMO equating SES with an email company is like saying home depot is a contractor because they sell hammers and lumber. It gives you the tools to be able to build stuff but it's not the same as a construction company.

SES, like Mailchimp, Sendgrid, and Postmark, has dedicated IPs as a paid add-on.

SES is not in the business of ensuring deliverability for your email. Dedicated ips is another hammer that they sell.

If Amazon can't make it work, might just use good old postfix I guess.

My anecdotal evidence is that simple postfix setup with IP from a reputable hoster which is not blacklisted works perfectly. And it's actually pretty simple to configure postfix for sending mail (now configuring postfix + dovecot + auth + spam filtering is another story).

> If Amazon can't make it work, might just use good old postfix I guess.

AWS blocks TCP25 outbound by default [1].

[1] https://repost.aws/knowledge-center/ec2-port-25-throttle

> My anecdotal evidence is that simple postfix setup with IP from a reputable hoster which is not blacklisted works perfectly.

And my anecdotal evidence is that it does not work perfectly: https://www.attejuvonen.fi/dont-send-email-from-your-own-ser...

As others mentioned they are using a paid AWS email service. In fairness however the price is very low thus making the bar to entry very low. They will have no shortage of customers that abuse the system, do not handle UCE reports and in some cases are outright spamming. The spammers may get blocked with time but there will be a continuous wake of damage in their path on the shared pools of IP addresses.

> you should be buying email delivery from a company who is in that business.

OP is doing just that, Amazon SES = Simple Email Service

Then OP should change services to a service that offers this as their primary mission objective.

If SES isn't delivering some stuff, Amazon will link to a page and say "here is why, good luck" and that's the end of the support ticket.

> OP is doing just that, Amazon SES = Simple Email Service

If you look at AWS pages for SES, nowhere on their pages do they market email deliverability. That's not the service they sell.

I think this is a spammy neighbour problem on that sending IP. Might be you, might be someone else who's using SES. But whoever sends from this IP is penalised.

Do you have a custom return-path configured? Using a custom return path might help, because it ties your reputation primarily to your domain.

ESPs check multiple factors. Both IP and domain reputation play a role. They will check your return path / envelope sender domain reputation and your IP. Your domain will start with it's own reputation, but can be boosted with a good IP reputation. But if your domain had bad sending behaviour in the past, that might be an issue.

Source: I'm running a transactional mail service that solely works with shared IPs: https://www.markix.com.

FYI: Your privacy policy is missing https://app.markix.com/privacy-policy

Amazon is very, very spammy. If you want your email to be delivered and you don't want to end up sharing Amazon's reputation, you'll need to use another company for email delivery.

>>> Is this the side effect of Gmail's dormant account deletion rolled out last week?

IIRC, they haven't started to actually close the accounts just yet, so I doubt it's related.

We are always trying to trim waste to stay lean, and the SES vs Sendgrid pricing looks nice on paper (we are on the Sendgrid Pro plan with the dedicated IP address, so it's $90/mo). However when I look at our Sendgrid stats (97% reputation; it's pretty much all transactional) I know it's worth well more than what we'd save.

Why does a 4xx even come to your attention? Admittedly I never looked into what SES is or how it works, but I assumed it stores and forwards messages on behalf of its users, in which case a 4xx temporary failure to send should not come back to you.

For those recommending against SES for critical deliveries, what are you using?

In the recent past I’ve used Postmark, but they were acquired by a marketing company.

For any critical email delivery make sure you can address this thing with the mailsender of choise:

1)SPF,

2)DKIM,

3)DMARC [2] (DMARC is often forgotten or can be super noisy when set up. Postmarc offers a aggregation service for free that sends you a weekly summary),

4)Dedicated IP

5)Reverse IP look up [1] (locate a dns PTR record for that IP address) should match the sender.

Not everyone supports 5).

4 and 5 is what you end up paying for, but totally worth it. Sendgrid, SES, Mailgun etc.

[1] https://www.mailgun.com/blog/deliverability/reverse-dns-whit...

[2] https://dmarc.postmarkapp.com/

And 1 surprising thing (to me) is some places (iCloud was what bit me) alway want you to be able to receive email (have MX records). Even though I have no need to support incoming emails iCloud was blocking my sending until I got that setup. The incoming emails just go to a black hole but that's enough.

Shameless plug: I've recently started my own transactional email service (https://www.markix.com), primarily targeting small senders, after having been a very happy Postmark customer for a long time. Our service is still in closed beta but delivering live emails.

I run a couple other businesses and moved all of my transactional email sending over to Markix.

Would love to have a chat with anyone that might be starting a new project and is open to try out a new mail service (mail in bio).

Funny because Gmail is the biggest source of spam

Gmail is also one of the biggest email providers... it's almost like there is a correlation.

I tried to use cloudflare email routing and had the same issue. I simply set it up so any email to @mydomain.com would forward to a gmail.

The worst is that cloudflare did not let me know this was happening until I saw I was missing some emails and went hunting. About 20% of my emails would just get rejected silently with "delivery failed" in the logs. I wouldnt blame cloudflare so much if they kept attempting to redeliver, but they did not. They simply give up.

Reply on news.ycombinator.com