Bug #15980 net_read_timeout set too low
Submitted: 26 Dec 2005 2:32 Modified: 9 Apr 2006 4:41
Reporter: Timothy Smith Email Updates:
Status: Closed Impact on me:
Category:Instance Manager Severity:S3 (Non-critical)
Version:5.0.17 OS:Any (any)
Assigned to: Jim Winstead CPU Architecture:Any

[26 Dec 2005 2:32] Timothy Smith
A connection to the instance manager times out after only 30 seconds.  Comments in server-tools/instance-manager/priv.cc suggest that the tool intends to copy the mysqld defaults:

unsigned long net_read_timeout= 30;             // same as in mysqld

However, mysqld defines several timeouts, and the one which should be copied is the NET_WAIT_TIMEOUT value, which is defined as 8 hours (8 * 60 * 60).

How to repeat:

Set up the instance manager.

Connect to it, via TCP or unix socket.  Issue a command, to verify it's working.  Wait 30 seconds.  Issue another command, and notice that your connection has timed out.

Suggested fix:

First, use the NET_WAIT_TIMEOUT macro for the default.

It might be appropriate to allow the user to set this (with wait_timeout and interactive_timeout) like mysqld.
[8 Mar 2006 17:50] Jim Winstead
Already fixed by Petr.
[31 Mar 2006 1:09] Timothy Smith
This was fixed on 10 Feb, 2006, in this changeset:


(Revision 1.10 to server-tools/instance-manager/priv.cc)

That means that it was fixed in 5.0.19 and 5.1.7.
[9 Apr 2006 4:41] Paul DuBois
Noted in 5.0.19, 5.1.7 changelogs.
See also Bug#12674.