Hi all, I'm running Postfix 2.11.0 on Ubuntu 14.04.1 LTS on multiple m3.xlarge instances (15GB RAM) on Amazon EC2. There's a milter plugged in. This setup has been running without problems on Postfix 2.9.6 on Ubuntu 12.04.2 LTS on bare metal machines (32GB RAM) for years. Only when we ported it to EC2 did we start seeing the following warnings:
> Dec 10 12:47:27 fbr-sample-00 postfix/smtpd[2350]: warning: connect to > private/tlsmgr: Resource temporarily unavailable > Dec 10 12:47:27 fbr-sample-00 postfix/smtpd[2350]: warning: problem talking > to server private/tlsmgr: Resource temporarily unavailable I did some digging in mailing list archives and on the web, and the most relevant reference I've been able to find was an old postfix-users thread saying this could be caused by a shortage of entropy from a blocking entropy source, such as /dev/random. However, Ubuntu's Postfix is compiled to use /dev/urandom, which is not supposed to block: > $ postconf tls_random_source > tls_random_source = dev:/dev/urandom Also, it seems like we have plenty of entropy available: > $ cat /proc/sys/kernel/random/entropy_avail > 877 I've never seen it dip below 750. It also seems like available RAM is not an issue. These machines use only a fraction of available RAM. What other possible causes are there for the above warnings? We suspect they are causing incoming SMTP transactions to be terminated and messages to be resent. Thanks, -Julian
