When I wrote Everything Needs Rate Limits, I mentioned in passing that the disk-full state prevented receiving email. The MTA was returning a temporary “try again later” code, but the clients weren’t responding. I got several session transcripts emailed to me that were of the form:
> EHLO some-host-name
< 200 OK + capabilities listing
> MAIL FROM sender-address
< 200 OK
> RCPT TO recipient-address
< 415 Resource unavailable, try again later
> RCPT TO recipient-address
< 415 Resource unavailable, try again later
> RCPT TO recipient-address
< 415 Resource unavailable, try again later
> DATA …
< 500 No recipient given
X Connection lost
The client is recognizing that they’re getting some sort of error, but their idea of “later” was milliseconds later, and the disk space problem was not being cleared at CPU speeds. After enough rejections of their recipient-address, they YOLO’d it and sent the email body, to no avail.
(I also think it’s pretty interesting that the MTA is happy to tell everyone else ENOSPC, yet deliver these error emails through to me. I suppose it purposely stops accepting email with some reserve disk space, so that it can continue to deliver critical errors for a while.)
No comments:
Post a Comment