Bug #28576 Slave recovery from a bad connection
Submitted: 21 May 2007 22:05 Modified: 23 May 2007 13:57
Reporter: Arjen Lentz Email Updates:
Status: Verified Impact on me:
None 
Category:MySQL Server: Replication Severity:S4 (Feature request)
Version:any OS:Any
Assigned to: Assigned Account CPU Architecture:Any
Tags: replication, SSL, tunnel, vpn
Triage: Triaged: D5 (Feature request)

[21 May 2007 22:05] Arjen Lentz
Description:
Replication can break due to connection problems (remote connections, tunnels) and sometimes this causes corruption in the binary log.

In theory of course, this shouldn't happen because the TCP link is error-corrected, but obviously it does anyway.

In many cases this will actually cause an SQL syntax error on the slave when it tries to execute the corrupted query. This is an absolute indication that something like the above has happened, because normally a query with an SQL error would not appear in the binlog.

How to repeat:
feature req. see description.

Suggested fix:
When such a specific situation occurs, the slave could try to seek back to that position on the master and re-read the event (just once).
This very likely fixes the problem, without replication stopping completely. Obviously, an error log warning should be issued anyway.
[23 May 2007 13:57] Miguel Solorzano
Thank you for the bug report feature request.