/ Zope / Apsis / Pound Mailing List / Archive / 2004 / 2004-08 / BIO_read() returns EAGAIN?

[ << ] [ >> ]

[ Pound on ntpl system / Thierry Coopman ... ] [ Flushing logmsg / Bart Teeuwisse ... ]

BIO_read() returns EAGAIN?
Bill Anderson <bill(at)interhack.com>
2004-08-13 00:24:53 [ FULL ]
I'm getting the following errors very often:

Aug 12 18:10:16 bobo pound: copy_bin premature EOF: Resource temporarily
unavailable
Aug 12 18:10:16 bobo pound: error copy server cont: Resource temporarily
unavailable

Much less frequently, I'm getting:

Aug 12 18:10:20 bobo pound: copy_bin error writing: Broken pipe
Aug 12 18:10:20 bobo pound: error copy server cont: Broken pipe


I tried setting "Server 30", in which case I get:

Aug 12 18:17:49 bobo pound: response error read from 127.0.0.1:8183: Unknown
error: 0


Previously my "Alive" value was unset.  Setting it didn't change anything.

I'm using pound 1.7 on FreeBSD 4.10.  Pound is between Apache (cache/ssl)
and several zope processes.  All communication between apache/pound/zope
is over the loopback (127.0.0.1) interface.

I did notice that one of my Zope processes seemed to have been getting
more load than the others.  Is my configuration correct to have pound
randomly select one of them?


Below is my pound.cfg:


# global stuff
ListenHTTP *,8180
User www
Group www
ExtendedHTTP 1
Alive 5
Client 20
Server 30

# zeo clients
UrlGroup ".*"
Backend 127.0.0.1,8181,1
Backend 127.0.0.1,8182,1
Backend 127.0.0.1,8183,1
Backend 127.0.0.1,8184,1
Session COOKIE _ZopeId 3600
EndGroup

MailBoxer