On Thu, 2005-07-14 at 12:09 +0800, Zhou Li wrote:
There are too many record like "(...http/client): Channel read timed out" in Zorp log, How can I disable recording the "Channel read timed out" message? any help will be appriciate
Zorp uses a 10sec timeout for the request line in keepalive mode and as browsers sometimes try to keep the connection open for a longer time Zorp times out and reports this in the error log. You can manipulate the way Zorp logs using the --verbose (-v) and --log-spec command line options. The message above is reported on core.error(1), this you can find out by using the --log-tags command line option in which case Zorp automatically inserts the associated logtag for each message. A log tag is comprised of a message class ('core.error' in the case above) and a verbosity level which controls what level is required for the message to actually be logged. Once you have a log tag you have two options: - change the global log verbosity using -v (default is 3), for the message above to disappear you'd need -v0, in which case you'd basically lose all logging - change the message class specific verbosity level using: -v3 --log-spec core.error:0 This way you leave the global verbosity level on 3, except for core.error messages where you set it to 0. For more information on logtags and the --log-spec command line option, see the manpage zorp(8). The problem with the solution I proposed above is that you probably don't want to lose complete error logging as other, more important problems are also logged this way. You could also try to increase the request timeout attribute in your proxy class: class LongTimeoutHttpProxy(HttpProxy): def config(self): HttpProxy.config(self) # request timeout in msec self.request_timeout = 20000 And use this class in your Service definition. -- Bazsi