On 07/16/2013 03:34 PM, Christian Rohmann wrote:
Hello Cluebringers,I've been running a git clone version from 16th of January (cdc44f58) for quota as well as for greylisting. Apart from the known bug in the IPv6 prefix definition (hint: "::/0") policydv2 appeared to be running fine. I found random occurrences of temporary configuration errors in the postfix logs. They correlate to these lines of policydv2 log: --- cut --- CBPOLICYD] WARNING: sysread[Connection reset by peer] => Peer: 127.0.0.1:33656, Local: 127.0.0.1:10031 --- cut --- The interesting bit is that there seems to be no other correlation. Policydv2 simply runs fine for hours at a time and then throws those kind of errors for a few minutes or hours and returns back to normal. There has never been any bypass switch or database timeout during that time. It's also different systems, different policies and different databases that show this exact behavior. Switching back to a much older 2.1 git version from Aug 31st 2011 (!) solves the issue, no more temporary errors or log lines mentioning sysread. Any clues or ideas?
The major difference between the above to versions is the commit of the pipelining fix...
http://devlabs.linuxassist.net/projects/policyd/repository/revisions/9da9a28d7c89df35d92737ab5e2e63507cf3db9dI can't see any other touching of the network code. The above bugfix is pretty important or you may be limited to under 10 mails per second.
Which version of postfix is this btw? the error indicates that policyd saw a connection reset from postfix, is it possible to strace one policyd process until this happens and log everything?
-N
smime.p7s
Description: S/MIME Cryptographic Signature
_______________________________________________ Users mailing list [email protected] http://lists.policyd.org/mailman/listinfo/users_lists.policyd.org
