On Wed, 2015-02-25 at 09:20 -0500, Alan Conway wrote: > ... > BUT it turns out the sporadic >1 second delays are coming from > CyrusSasl::start!!? (see logs below) Can anybody tell my why that would > ever take more than 1000ms for (ANONYMOUS PLAIN) mechanisms that involve > 0 encryption? Even the successful connect takes 300ms. I know this isn't > on the critical path but, wow. There was a time we could do <6ms latency > on clustered qpidd. Seems to me that's a bit more complex than saying > "no idea who you are, carry on!"
It may be worth checking whether this time is spent in our code (qpid code) or the cyrus sasl code. My guess is that this time is spent in sasl_client_new() or sasl_client_init() setting up cyrus sasl to process this connection rather than sasl_client_start() actually processing the mech list which as you say does not have to do very much. However if I remember correctly the other 2 are also run at the same time and they may have more significant set up work to do (even if it isn't relevant for this actual connection) Andrew --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
