MySQL Bugs: #24826: Patch: number of connections to reserved for SUPER users

Bug #24826	Patch: number of connections to reserved for SUPER users
Submitted:	5 Dec 2006 16:55	Modified:	7 Feb 2007 17:14
Reporter:	Lenz Grimmer	Email Updates:
Status:	In progress	Impact on me:	None
Category:	MySQL Server: General	Severity:	S4 (Feature request)
Version:	5.0-community	OS:	Any
Assigned to:	Assigned Account	CPU Architecture:	Any
Tags:	Contribution
Triage:	Triaged: D5 (Feature request)

Description:
This bug report is a placeholder to keep track of the progress of including this
patch from Jeremy Cole to display the uptime since the last FLUSH STATUS
command, available from http://jcole.us/patches/mysql/5.0/reserved_connections.patch

How to repeat:
Apply the patch

Suggested fix:
Apply the patch

A patch for this bug has been committed. After review, it may
be pushed to the relevant source trees for release in the next
version. You can access the patch from:

  http://lists.mysql.com/commits/19048

ChangeSet@1.2359, 2007-01-30 17:00:14-05:00, cmiller@zippy.cornsilk.net +4 -0
  Bug#24826: Number of connections to reserved for SUPER users
  
  There are two sites that control the number of connections that we
  support.  The network level is willing to listen for N + 1 
  connections.  The access control level is willing to accept N 
  connections.  In this way, we can answer a connection, tell it
  "sorry" if it's a normal user and drop it.  In this way, we reserve
  space for one superuser connection at present.
  
  This patch increases the number of network connections we're willing
  to support, above the N normal users.

Any reason *why* we could want to apply this (besides "a contributor sent it") ?

for +10 there's a comment in mysqld.cc:

  /*
    Setup alarm handler
    This should actually be '+ max_number_of_slaves' instead of +10,
    but the +10 should be quite safe.
  */

The "10" should have a define that is documented (don't use magic numbers in the server).

To answer why we want this: you go in as super, kill a thread. It's now stuck in a ten minute rollback and you can't do anything else to solve the load problem that has caused all connections to be used. More super connections gives you a chance to get more done and reduces the risk associated with the actions you take. You can also find that automated tools have consumed the super connection that you need to use to intervene.