From: Michael Dykman Date: October 11 2012 4:59pm Subject: Re: RE: innodb_lock_wait_timeout List-Archive: http://lists.mysql.com/mysql/228373 Message-Id: MIME-Version: 1.0 Content-Type: multipart/alternative; boundary=f46d042e00c130ae9e04cbcb7fb3 --f46d042e00c130ae9e04cbcb7fb3 Content-Type: text/plain; charset=ISO-8859-1 The original poster mentioned that he is not using transactions explicitly. Some transactions may still occur as a side effect of some operations under certain conditions and, in a busy high load environment, cannot be entirely avoided. Having some experience with this, I can report that it is safe and highly effective to retry at the application layer. With a site supporting 1.5M users/day, we set a loop to retry up to 3 times.. Out of tens of millions of writes/day, we only hit the lock/timeout a couple hundred times, and never needed the second retry. On 2012-10-11 12:36 PM, "Rick James" wrote: A 50-second 'transaction' is much too long. If you have slow queries, let's see them, together with SHOW CREATE TABLE, SHOW TABLE STATUS, and EXPLAIN SELECT. Quite possibly we can make them run faster, thereby eliminating your problem. > -----Original Message----- > From: Akshay Suryavanshi [mailto:akshay.suryavanshi50@stripped] >... --f46d042e00c130ae9e04cbcb7fb3--