qmail smtp server no longer starts up using latest kernel
| Affects | Status | Importance | Assigned to | Milestone | |
|---|---|---|---|---|---|
| qmail (Ubuntu) |
New
|
Undecided
|
Unassigned | ||
Bug Description
Following on from some patching I've done this week after the release of the latest kernels, I observed that qmail-smtpd no longer starts due to the following error:
"tcpserver: fatal: unable to figure out port number for smtp"
You will see this is logged repeatedly out to "/var/log/
I've just spent a little time tracing it and have concluded that the softlimit defined in "/etc/service/
Changing the softlimit from 7000000 to 9000000 allows the service to start up as before, as does rolling back to the previous kernel. Presumably the stack clash patching has caused this change in memory requirements for the process and hence the softlimit increase.
I've just provisioned a clean Ubuntu 16.04 host to confirm that it's not any existing configuration:
* Installed Ubuntu 16.04 on Digital ocean
* edit /etc/hosts to ensure a FQDN is present
* apt-get update
* apt-get install qmail
* qmailctl stat
"/etc/
* tail /var/log/
"@4000000059
I've observed the same issue on hosts running Ubuntu 16.04, 14.04 and Debian 8 for what it's worth.
| Wilb (ubuntu-wilb) wrote | #1 |
| Wilb (ubuntu-wilb) wrote | #2 |
Apologies, I missed a couple of lines from my steps to reproduce when copy & pasting it over from my terminal history.
It should read something more like:
* Installed Ubuntu 16.04 on Digital ocean
* edit /etc/hosts to ensure a FQDN is present
* apt-get update
* apt-get install qmail
* qmailctl stat & tail -1 /var/log/
"/etc/
"tcpserver: fatal: unable to figure out port number for smtp"
* sed -i s/7000000/900000/ /etc/service/
* tail -1 /var/log/
"status: local 0/10 remote 0/20"