On 9/08/2013 2:04 p.m., Romeo Mihalcea wrote:
I just upgraded to squid 3.3.8 to take advantage of the workers and split the load between CPUs. Since I never worked with workers and the documentation or examples are lacking I thought I start by simply adding workers 2 to my config.
http://wiki.squid-cache.org/Features/SmpScale is pretty comprehensive for the unless you are interested in coding on the SMP features, even this it is the required reading.
Pay attention to the "What_can_workers_share?" section details, and "Troubleshooting" and check of any of the known issues are relevant to your proxy before using workers in production. Most of them have workarounds but we are not widely publishing those due to the rather fluid nature of the config hacks requried. A post here asking about particualr issues will usually get a response.
Squid start with no error but the clients can't connect to the http_port. If I remove the workers line everything works allright. Anything else I should change to have it go? P.S. I changed nothing else from a working config. Just added: workers 2
Do you have a very large cache? if the workers take too long to load they may have trouble registering with the coordinator process to retrieve the listening socket details.
cache.log might have more info to assist with debugging. Amos
