Some people have been reporting problems with mail to gmail backing up. Steve has some information about the problem.
I’m seeing comments in a few places that some gmail MXes (but not all) are tempfailing some pipelined SMTP transactions, which is causing mail being sent there by some smarthosts to back up.
I don’t have details, but if you’re seeing “451 4.5.0 SMTP protocol violation, see RFC 2821” that may be your problem. If you’re running a recent-ish (2.2 or later) version of postfix then “postconf -e smtp_discard_ehlo_keywords=pipelining” followed by “postfix reload” will turn off pipelining altogether as a temporary workaround.
It appears right now to just affecting Postfix, but Steve later says:
I don’t have enough details to know for sure, but I suspect it’ll be hitting most, maybe all, MTAs that use SMTP pipelining at all aggressively. I’d expect most MTAs used by ESPs to be reasonably aggressive with pipelining, as it reduces latency and ups performance a lot.
So it’s definitely worth checking your queues and tempfail logs.
If you’re seeing queues backed up to gmail or you’re seeing that particular 451 rejections from gmail then your smarthost has the problem. If you’re not seeing any problems, it probably doesn’t.
RSS - Posts
We’re having this exact same problem….where have Steve, or others, been tracking this issue?
Hi, Frank,
My understanding is that it was mentioned on the mailop list and that post said that the issue was being tracked on the Postfix list.
http://archives.neohapsis.com/archives/postfix/2008-11/0166.html is the mention on the postfix list. There’s no suggestion it’s postfix-specific, but that list is where I saw it mentioned (and I mostly trust postfix to implement ESMTP correctly, so if a postfix box can’t interoperate with gmail, I’ll tend to blame gmail).
I was chatting privately with someone with the same problem on a couple of servers. Disabling pipelining seems to have resolved it, so it looks likely that Google broke pipelining support on at least some of their inbound cluster.
I’ve also heard mumblings from several places that quite a few other servers are seeing the same issue, and the common factor is mailing to gmail.
Thanks for those confirming comments and pointing to the source of the postfix tweaks.
Gmail may have fixed their issue because my outbound queues don’t have any gmail.com messages in there.
The mailops listserv also clarified that it’s pipelining support that is/was the problem and not postfix.
I can confirm that was a problem on Google side and also they fixed the problem. I was suffering big time yesterday with all servers queues full with ‘protocol compliance’ rejection messages. Disabling pipelining cleaned the queues. Today I tried to re-enable pipelining and delivery to Google remained ok.
Not sure if this is the right place to drop my question? I frequently have mailerdemom return mail that I have sent, telling me I’m using an incorrect email address for the addressee. However when I Forward my initial email to the same addressee, using the same email address the email is received.
Sorry if I violated protocol here, but AOL doesn’t publish a an address for online help for non-subscribers.
I use po3 Verizon, and Mozilla’s Thunderbird. I can not receive any incoming gmail. There doesn’t seem to be any blocks.